handle
stringclasses 20
values | did
stringclasses 20
values | uri
stringlengths 56
70
| cid
stringlengths 59
59
| text
stringlengths 60
1.48k
| originalText
stringlengths 1
1.4k
| replyCount
int64 0
5
| createdAt
stringdate 2009-10-15 18:56:15
2025-06-21 21:20:01
| indexedAt
stringdate 2009-10-15 18:56:15
2025-06-21 21:20:04
|
|---|---|---|---|---|---|---|---|---|
emollick.bsky.social
|
did:plc:flxq4uyjfotciovpw3x3fxnu
|
at://did:plc:flxq4uyjfotciovpw3x3fxnu/app.bsky.feed.post/3lgslb6iqis2l
|
bafyreifzw5q4x7tfohbsgevi74lgy6xws7xd2xm5a35vbyfn6vbluo5v6m
|
<start>Razors are philosophic tools for simplifying decisions.
“Claude, come up with some clever razors.” A few are really solid.
@emollick.bsky.social
|
Razors are philosophic tools for simplifying decisions.
“Claude, come up with some clever razors.” A few are really solid.
| 0 |
2025-01-28T13:38:18.769Z
|
2025-01-28T13:38:20.326Z
|
emollick.bsky.social
|
did:plc:flxq4uyjfotciovpw3x3fxnu
|
at://did:plc:flxq4uyjfotciovpw3x3fxnu/app.bsky.feed.post/3lgridgsal22g
|
bafyreiap5djg65vubcp7v2dwxn3epd6am2lesoghflnxt7x3i6j32go3xy
|
<start>Running DeepSeek r1 32B locally is kind of depressing. You get the reasoning, but the model is not *smart* enough to actually use it so it keeps getting tripped up in its own ideas & is full of doubt.
3 minutes on a limerick that a small non-reasoning model could do instantly.
@emollick.bsky.social
|
Running DeepSeek r1 32B locally is kind of depressing. You get the reasoning, but the model is not *smart* enough to actually use it so it keeps getting tripped up in its own ideas & is full of doubt.
3 minutes on a limerick that a small non-reasoning model could do instantly.
| 0 |
2025-01-28T03:13:13.614Z
|
2025-01-28T03:13:15.419Z
|
emollick.bsky.social
|
did:plc:flxq4uyjfotciovpw3x3fxnu
|
at://did:plc:flxq4uyjfotciovpw3x3fxnu/app.bsky.feed.post/3lgreswpbhc2g
|
bafyreiamgexnc67pvbtgavzichnu73lqhuijni57hl3lpto6wrqlv6vaja
|
<start>DeepSeek r1 7B versus Llama 3 8B on my PC...
I don't think it is worth running small versions of DeepSeek r1 locally yet (but obviously early days).
@emollick.bsky.social
|
DeepSeek r1 7B versus Llama 3 8B on my PC...
I don't think it is worth running small versions of DeepSeek r1 locally yet (but obviously early days).
| 0 |
2025-01-28T02:10:18.647Z
|
2025-01-28T02:10:21.218Z
|
emollick.bsky.social
|
did:plc:flxq4uyjfotciovpw3x3fxnu
|
at://did:plc:flxq4uyjfotciovpw3x3fxnu/app.bsky.feed.post/3lgqlv36l522b
|
bafyreifyex2tws4lvmjb5itkhv7wr2bc4sgugq7wztpqskiiqvem4ums64
|
<start>Deleted a post on Janus 7B because it is unfair to compare a diffusion model to a multimodal model.
Its not a great image generator compared to small diffusion models👇, but until we find out parameters from something like Aurora, it is hard to compare
@emollick.bsky.social
|
Deleted a post on Janus 7B because it is unfair to compare a diffusion model to a multimodal model.
Its not a great image generator compared to small diffusion models👇, but until we find out parameters from something like Aurora, it is hard to compare
| 0 |
2025-01-27T18:44:06.905Z
|
2025-01-27T18:44:08.722Z
|
emollick.bsky.social
|
did:plc:flxq4uyjfotciovpw3x3fxnu
|
at://did:plc:flxq4uyjfotciovpw3x3fxnu/app.bsky.feed.post/3lgqi2d36bk2r
|
bafyreiarvcjiicta2nc343jqg4cl5mpkvmxjqljsaf7yi3qaicqf2pma6a
|
<start>In games used in behavioral economics to test strategic reasoning, OpenAI’s o1 “consistently outperforms other LLMs and human subjects.”
Reasoning models really do seem to offer fundamentally different higher-order capabilities compared to other AIs. arxiv.org/pdf/2412.13013
@emollick.bsky.social
|
In games used in behavioral economics to test strategic reasoning, OpenAI’s o1 “consistently outperforms other LLMs and human subjects.”
Reasoning models really do seem to offer fundamentally different higher-order capabilities compared to other AIs. arxiv.org/pdf/2412.13013
| 0 |
2025-01-27T17:35:27.978Z
|
2025-01-27T17:35:29.829Z
|
emollick.bsky.social
|
did:plc:flxq4uyjfotciovpw3x3fxnu
|
at://did:plc:flxq4uyjfotciovpw3x3fxnu/app.bsky.feed.post/3lgq37hb3xs2g
|
bafyreiablpyo74k6ce6bpnc4a6hsa4ftutiisu7ekitigkndij3oysnaym
|
<start>This weekend I wrote a post on which AI to use right now (at least for general, individual users). Model strength may matter less to most users than the capabilities of the apps and the other features that each model includes. It is a little complicated.
www.oneusefulthing.org/p/which-ai-t...
@emollick.bsky.social
|
This weekend I wrote a post on which AI to use right now (at least for general, individual users). Model strength may matter less to most users than the capabilities of the apps and the other features that each model includes. It is a little complicated.
www.oneusefulthing.org/p/which-ai-t...
| 0 |
2025-01-27T13:45:41.507Z
|
2025-01-27T13:45:44.119Z
|
emollick.bsky.social
|
did:plc:flxq4uyjfotciovpw3x3fxnu
|
at://did:plc:flxq4uyjfotciovpw3x3fxnu/app.bsky.feed.post/3lgp4pnujac2p
|
bafyreifvjzo4brnnkj2qy4rforyiixtyxmf7jpxaprph2q6brf4yckffni
|
<start>From historian Prof. Ben Breen: “yes, these things [AI] can definitely “do” historical research and analysis now, and I am 100% certain that they will improve many aspects of the work historians do to understand the past, especially in the realms of transcription, translation & image analysis” open.substack.com/pub/resobscu...
@emollick.bsky.social
|
From historian Prof. Ben Breen: “yes, these things [AI] can definitely “do” historical research and analysis now, and I am 100% certain that they will improve many aspects of the work historians do to understand the past, especially in the realms of transcription, translation & image analysis”
| 1 |
2025-01-27T04:39:59.301Z
|
2025-01-27T04:40:00.822Z
|
emollick.bsky.social
|
did:plc:flxq4uyjfotciovpw3x3fxnu
|
at://did:plc:flxq4uyjfotciovpw3x3fxnu/app.bsky.feed.post/3lgog4li4as2g
|
bafyreia7ryevdrbh7eouw6cbenqkv3pbartofgzrjk66zi5ovgvaazfnoi
|
<start>To be clear, is mostly something that matters in data analysis, where human raters are needed, but also expensive and unreliable.
@emollick.bsky.social
|
To be clear, is mostly something that matters in data analysis, where human raters are needed, but also expensive and unreliable.
| 0 |
2025-01-26T21:55:36.952Z
|
2025-01-26T21:55:38.527Z
|
emollick.bsky.social
|
did:plc:flxq4uyjfotciovpw3x3fxnu
|
at://did:plc:flxq4uyjfotciovpw3x3fxnu/app.bsky.feed.post/3lgofqfwzgs2g
|
bafyreidum227zlw3vbalu76c4m4mug7a6woli4h6hgepwjp7m5rvv7znyi
|
<start>Nice way to test when AI can replace human evaluators & judges, it compares if LLMs align better with group consensus than individual human evaluators do
GPT-4 & Gemini pass the test in 8/10 tasks, but struggle with deep contextual understanding. Few-shot learning helps. arxiv.org/abs/2501.10970 To be clear, is mostly something that matters in data analysis, where human raters are needed, but also expensive and unreliable.
@emollick.bsky.social
|
Nice way to test when AI can replace human evaluators & judges, it compares if LLMs align better with group consensus than individual human evaluators do
GPT-4 & Gemini pass the test in 8/10 tasks, but struggle with deep contextual understanding. Few-shot learning helps. arxiv.org/abs/2501.10970
| 1 |
2025-01-26T21:48:48.492Z
|
2025-01-26T21:48:50.825Z
|
emollick.bsky.social
|
did:plc:flxq4uyjfotciovpw3x3fxnu
|
at://did:plc:flxq4uyjfotciovpw3x3fxnu/app.bsky.feed.post/3lgnlymhs3226
|
bafyreifji7e5hn56xpswiaph436lck5lqavqggs6al7xqc6s5mbzhzj66i
|
<start>Every few months, I write an opinionated guide for general purpose users about which AI to pick, especially for newcomers.
Here is my brand new one, which I actually had to update multiple times in the few days I was writing it. Things are changing fast. open.substack.com/pub/oneusefu...
@emollick.bsky.social
|
Every few months, I write an opinionated guide for general purpose users about which AI to pick, especially for newcomers.
Here is my brand new one, which I actually had to update multiple times in the few days I was writing it. Things are changing fast. open.substack.com/pub/oneusefu...
| 0 |
2025-01-26T14:08:06.481Z
|
2025-01-26T14:08:08.015Z
|
emollick.bsky.social
|
did:plc:flxq4uyjfotciovpw3x3fxnu
|
at://did:plc:flxq4uyjfotciovpw3x3fxnu/app.bsky.feed.post/3lgmmuw2dvs2k
|
bafyreibcna6bvpwd62spbd5zjv3twubipspci3434ymna4sgssvct47i3i
|
<start>"o1, procedurally generated game, make it fun and original"
"this is really boring and typical. i asked for something truly original"
"an infinite runner alone is not startlingly original"
It made this platformer with color switching, all its own decision-making (I pointed out two bugs)
@emollick.bsky.social
|
"o1, procedurally generated game, make it fun and original"
"this is really boring and typical. i asked for something truly original"
"an infinite runner alone is not startlingly original"
It made this platformer with color switching, all its own decision-making (I pointed out two bugs)
| 0 |
2025-01-26T04:51:16.316Z
|
2025-01-26T04:51:17.215Z
|
emollick.bsky.social
|
did:plc:flxq4uyjfotciovpw3x3fxnu
|
at://did:plc:flxq4uyjfotciovpw3x3fxnu/app.bsky.feed.post/3lglxbmgopc2e
|
bafyreib7ppe6d3xf2d5yfwf5q7s2xtifq5toh46jyjvcrmxmtkwr5lvvse
|
<start>This is a big concern for ne as well, especially on this site.
AI has many limitations, and there are lots of things to criticize about AI and lots to be concerned about (as well as lots of things that are very exciting and very positive for the world), but it isn’t fake or going away.
@emollick.bsky.social
|
This is a big concern for ne as well, especially on this site.
AI has many limitations, and there are lots of things to criticize about AI and lots to be concerned about (as well as lots of things that are very exciting and very positive for the world), but it isn’t fake or going away.
| 0 |
2025-01-25T22:24:40.118Z
|
2025-01-25T22:24:40.224Z
|
emollick.bsky.social
|
did:plc:flxq4uyjfotciovpw3x3fxnu
|
at://did:plc:flxq4uyjfotciovpw3x3fxnu/app.bsky.feed.post/3lglj5o6oxc2b
|
bafyreigm4goziqxuj6hhbq6nc4ufj6g2vy7h6y37paf5pyy64apvmwjyyq
|
<start>The broader picture of the last month is not DeepSeek or StarGate or whatever but that all the technical & investment signals are now pointing to an expectation of continued rapid acceleration of AI capabilities.
And we still lack a clear articulation about what that looks like from any major party
@emollick.bsky.social
|
The broader picture of the last month is not DeepSeek or StarGate or whatever but that all the technical & investment signals are now pointing to an expectation of continued rapid acceleration of AI capabilities.
And we still lack a clear articulation about what that looks like from any major party
| 0 |
2025-01-25T18:11:55.354Z
|
2025-01-25T18:11:55.524Z
|
emollick.bsky.social
|
did:plc:flxq4uyjfotciovpw3x3fxnu
|
at://did:plc:flxq4uyjfotciovpw3x3fxnu/app.bsky.feed.post/3lgkbyk2vhs2k
|
bafyreifpnf3rnmydaoya4uc3wfn4zl3oxfknpgr55vlnkcivh73qiggoeq
|
<start>It isn't clear how to use it if you don't know it is there, but you have to select Canvas from the little tool icon under the prompt for o1 to realize it can use Canvas, and you have to hit the "preview" icon afterwards, otherwise it just gives you code. Code for those who asked: docs.google.com/document/d/1...
@emollick.bsky.social
|
It isn't clear how to use it if you don't know it is there, but you have to select Canvas from the little tool icon under the prompt for o1 to realize it can use Canvas, and you have to hit the "preview" icon afterwards, otherwise it just gives you code.
| 1 |
2025-01-25T06:31:07.332Z
|
2025-01-25T06:31:07.619Z
|
emollick.bsky.social
|
did:plc:flxq4uyjfotciovpw3x3fxnu
|
at://did:plc:flxq4uyjfotciovpw3x3fxnu/app.bsky.feed.post/3lgkbosznks2k
|
bafyreif4wso2rwz7g62xf7gjv45vaix2iesbb74z4akg2tcxhk5mmg4jmu
|
<start>Coding by just asking for you want really feels magical. The advantage over sonnet is that o1 seems to maintain context better, the code can be much longer, and there are fewer constraints on what it can run. It isn't clear how to use it if you don't know it is there, but you have to select Canvas from the little tool icon under the prompt for o1 to realize it can use Canvas, and you have to hit the "preview" icon afterwards, otherwise it just gives you code.
@emollick.bsky.social
|
Coding by just asking for you want really feels magical. The advantage over sonnet is that o1 seems to maintain context better, the code can be much longer, and there are fewer constraints on what it can run.
| 1 |
2025-01-25T06:25:41.183Z
|
2025-01-25T06:25:41.519Z
|
emollick.bsky.social
|
did:plc:flxq4uyjfotciovpw3x3fxnu
|
at://did:plc:flxq4uyjfotciovpw3x3fxnu/app.bsky.feed.post/3lgkblryxxc2k
|
bafyreidnfa2zfwxwftyduu5v2buishlfcc2bdnlvqepd23kd6hwcavkdz4
|
<start>The new o1 canvas is quite impressive at coding from intent: "create an interactive tool that visually shows me how correlation works, and why correlation alone is not a great descriptor of the underlying data in many cases. make it accessible to nonmath people and highly interactive and engaging" Coding by just asking for you want really feels magical. The advantage over sonnet is that o1 seems to maintain context better, the code can be much longer, and there are fewer constraints on what it can run.
@emollick.bsky.social
|
The new o1 canvas is quite impressive at coding from intent: "create an interactive tool that visually shows me how correlation works, and why correlation alone is not a great descriptor of the underlying data in many cases. make it accessible to nonmath people and highly interactive and engaging"
| 1 |
2025-01-25T06:23:59.450Z
|
2025-01-25T06:24:00.218Z
|
emollick.bsky.social
|
did:plc:flxq4uyjfotciovpw3x3fxnu
|
at://did:plc:flxq4uyjfotciovpw3x3fxnu/app.bsky.feed.post/3lgj2jmyafk2k
|
bafyreiff5zems3tvfwsnjqmoceqiqxehx2oafkqdwikvtzy7vahph7bmwa
|
<start>People are likely way underestimating the true cost of training DeepSeek r1 (which does not make it less of an achievement, but it isn't just a small group of people who spent $5.5M) substack.com/home/post/p-...
@emollick.bsky.social
|
People are likely way underestimating the true cost of training DeepSeek r1 (which does not make it less of an achievement, but it isn't just a small group of people who spent $5.5M) substack.com/home/post/p-...
| 0 |
2025-01-24T18:44:51.141Z
|
2025-01-24T18:44:53.525Z
|
emollick.bsky.social
|
did:plc:flxq4uyjfotciovpw3x3fxnu
|
at://did:plc:flxq4uyjfotciovpw3x3fxnu/app.bsky.feed.post/3lgixed7n4c2a
|
bafyreih4ri33llty3jivaascnyd6l7llnq24bhft24gihla6oiyj2ahlmm
|
<start>Operator does a pretty good job in building a Magic deck using an online deck-building site.
The serious point is that using the web as a tool can help with error checking and act as a source of truth that keeps the LLM on track. For example, the website keeps track of card count, a LLM weakness.
@emollick.bsky.social
|
Operator does a pretty good job in building a Magic deck using an online deck-building site.
The serious point is that using the web as a tool can help with error checking and act as a source of truth that keeps the LLM on track. For example, the website keeps track of card count, a LLM weakness.
| 0 |
2025-01-24T17:48:11.900Z
|
2025-01-24T17:48:14.632Z
|
emollick.bsky.social
|
did:plc:flxq4uyjfotciovpw3x3fxnu
|
at://did:plc:flxq4uyjfotciovpw3x3fxnu/app.bsky.feed.post/3lgispxsao22a
|
bafyreih7fn3vjijkylfthwo2c6f4ik7zt72mlqgrrizfghluvjhkexrdha
|
<start>I suspect it is going to get increasingly tough to criticize or compare popular AIs on social media because people are increasingly getting somewhat emotionally connected to their model of choice. You aren’t critiquing model ability, you are insulting their helpful buddy.
@emollick.bsky.social
|
I suspect it is going to get increasingly tough to criticize or compare popular AIs on social media because people are increasingly getting somewhat emotionally connected to their model of choice. You aren’t critiquing model ability, you are insulting their helpful buddy.
| 0 |
2025-01-24T16:25:13.873Z
|
2025-01-24T16:25:14.626Z
|
emollick.bsky.social
|
did:plc:flxq4uyjfotciovpw3x3fxnu
|
at://did:plc:flxq4uyjfotciovpw3x3fxnu/app.bsky.feed.post/3lgikejhxxk23
|
bafyreidijx4da5az2lcl2icfv7bggmsghtuu7gk2aa2ocoraze4msnjqeq
|
<start>This is a big deal - the first open model (and non-US model) to get near the top of the AI leaderboard. While all benchmarks have flaws, the LM arena is an important one based on head-to-head comparisons by many users.
@emollick.bsky.social
|
This is a big deal - the first open model (and non-US model) to get near the top of the AI leaderboard. While all benchmarks have flaws, the LM arena is an important one based on head-to-head comparisons by many users.
| 0 |
2025-01-24T13:55:39.811Z
|
2025-01-24T13:55:41.624Z
|
emollick.bsky.social
|
did:plc:flxq4uyjfotciovpw3x3fxnu
|
at://did:plc:flxq4uyjfotciovpw3x3fxnu/app.bsky.feed.post/3lghmgx6ss22a
|
bafyreib3qvwjnvzdiekecqjvyzorsskw5c2jenst7q4cy55rswzwox7ceq
|
<start>"now draw what passes for the flag of a nation in that world"
Honestly, Claude's animated flag and explanation can't be beat: "The Resonance Banner serves as both a national symbol and a functional magical instrument - it's not just a flag, but a dynamic enchanted construct. (it goes on)" Claude explains
@emollick.bsky.social
|
"now draw what passes for the flag of a nation in that world"
Honestly, Claude's animated flag and explanation can't be beat: "The Resonance Banner serves as both a national symbol and a functional magical instrument - it's not just a flag, but a dynamic enchanted construct. (it goes on)"
| 1 |
2025-01-24T05:00:09.057Z
|
2025-01-24T05:00:10.218Z
|
emollick.bsky.social
|
did:plc:flxq4uyjfotciovpw3x3fxnu
|
at://did:plc:flxq4uyjfotciovpw3x3fxnu/app.bsky.feed.post/3lghlvvjowk2a
|
bafyreifiegehs2tlavilwsphcixpog67rxvsckx5q44tsdwy2zw3amu65i
|
<start>"Hey Claude, Gemini, o1 & DeepSeek, create an entirely new kind of fantasy map for a fantasy realm (like a Mercator projection or Polynesian tidal maps). the map should adhere to conventions that have in-world justifications, and be coherent and understandable. Show me an svg of the map and explain" "now draw what passes for the flag of a nation in that world"
Honestly, Claude's animated flag and explanation can't be beat: "The Resonance Banner serves as both a national symbol and a functional magical instrument - it's not just a flag, but a dynamic enchanted construct. (it goes on)"
@emollick.bsky.social
|
"Hey Claude, Gemini, o1 & DeepSeek, create an entirely new kind of fantasy map for a fantasy realm (like a Mercator projection or Polynesian tidal maps). the map should adhere to conventions that have in-world justifications, and be coherent and understandable. Show me an svg of the map and explain"
| 1 |
2025-01-24T04:50:36.876Z
|
2025-01-24T04:50:39.621Z
|
emollick.bsky.social
|
did:plc:flxq4uyjfotciovpw3x3fxnu
|
at://did:plc:flxq4uyjfotciovpw3x3fxnu/app.bsky.feed.post/3lghdjqkmuk2e
|
bafyreicwwos24bqgjvgin3oikmza67w64p7hi2ugivxfx2yp2nmbplfmai
|
<start>Oh god I think I helped accidentally kick off an new SEO industry by also posting this on LinkedIn. Sorry.
@emollick.bsky.social
|
Oh god I think I helped accidentally kick off an new SEO industry by also posting this on LinkedIn. Sorry.
| 0 |
2025-01-24T02:20:39.090Z
|
2025-01-24T02:20:39.622Z
|
emollick.bsky.social
|
did:plc:flxq4uyjfotciovpw3x3fxnu
|
at://did:plc:flxq4uyjfotciovpw3x3fxnu/app.bsky.feed.post/3lghdf3ggvs2e
|
bafyreib2we762m64yixnehguwxhr5xrcjverb7bemae3r2ip5yegk4u6je
|
<start>Next big thing for brands: knowing what sites agents prefer.
If you ask for stock prices, Claude with Computer Use goes to Yahoo Finance while Operator does a Bing search
Operator loves buying from the top search result on Bing. Claude has direct preferences like 1-800-Flowers
We don't know why Oh god I think I helped accidentally kick off an new SEO industry by also posting this on LinkedIn. Sorry.
@emollick.bsky.social
|
Next big thing for brands: knowing what sites agents prefer.
If you ask for stock prices, Claude with Computer Use goes to Yahoo Finance while Operator does a Bing search
Operator loves buying from the top search result on Bing. Claude has direct preferences like 1-800-Flowers
We don't know why
| 1 |
2025-01-24T02:18:02.713Z
|
2025-01-24T02:18:04.928Z
|
emollick.bsky.social
|
did:plc:flxq4uyjfotciovpw3x3fxnu
|
at://did:plc:flxq4uyjfotciovpw3x3fxnu/app.bsky.feed.post/3lggq7w2mok2e
|
bafyreid6qo3mhr3bethsdmrngxx5ivionaoz4oebx7ofsuet7jpve2fhu4
|
<start>Played with the new Operator for a little bit before launch:
1) General purpose web agents aren’t there yet, but seem more workable than expected(Operator is quite good)
2) Companies aren’t thinking enough about how to market to the preferences of agents
3) Security is going to get very weird fast
@emollick.bsky.social
|
Played with the new Operator for a little bit before launch:
1) General purpose web agents aren’t there yet, but seem more workable than expected(Operator is quite good)
2) Companies aren’t thinking enough about how to market to the preferences of agents
3) Security is going to get very weird fast
| 0 |
2025-01-23T20:35:08.215Z
|
2025-01-23T20:35:10.025Z
|
emollick.bsky.social
|
did:plc:flxq4uyjfotciovpw3x3fxnu
|
at://did:plc:flxq4uyjfotciovpw3x3fxnu/app.bsky.feed.post/3lgedhutnm22z
|
bafyreifstaqed5pbgdokxvad6yn3vhxbng4w72anvc5z2efaeiafaqcshu
|
<start>Sure, and that is a completely valid reason to quit X, among many others.
@emollick.bsky.social
|
Sure, and that is a completely valid reason to quit X, among many others.
| 0 |
2025-01-22T21:41:37.257Z
|
2025-01-22T21:41:39.519Z
|
emollick.bsky.social
|
did:plc:flxq4uyjfotciovpw3x3fxnu
|
at://did:plc:flxq4uyjfotciovpw3x3fxnu/app.bsky.feed.post/3lgedegszh22z
|
bafyreicdwurr2padmtdsa3j3mnq5uqyq6btnsrgy6cthjukahk274x7xla
|
<start>I don’t think that is a reasonable broad statement though it is true in some cases. I think that online communities are remarkably durable and many people don’t spend a lot of time thinking about whether Zuckerberg or Musk or Dorsey owns the platform that is useful for them.
@emollick.bsky.social
|
I don’t think that is a reasonable broad statement though it is true in some cases. I think that online communities are remarkably durable and many people don’t spend a lot of time thinking about whether Zuckerberg or Musk or Dorsey owns the platform that is useful for them.
| 0 |
2025-01-22T21:39:41.892Z
|
2025-01-22T21:39:43.024Z
|
emollick.bsky.social
|
did:plc:flxq4uyjfotciovpw3x3fxnu
|
at://did:plc:flxq4uyjfotciovpw3x3fxnu/app.bsky.feed.post/3lgebxomwx22z
|
bafyreigd7l33ndeeql5ptqxhrf7nujm3wb46e2rvek2vlrl7zynw5v5v7m
|
<start>When people ask about the image generator they should use, I point them at Google’s ImageFX for Imagen 3.
Midjourney is better for art, Recraft is better for workflows, etc. but Imagen 3 is just the easiest to use and most likely to produce what you want the first time. labs.google/fx/tools/ima...
@emollick.bsky.social
|
When people ask about the image generator they should use, I point them at Google’s ImageFX for Imagen 3.
Midjourney is better for art, Recraft is better for workflows, etc. but Imagen 3 is just the easiest to use and most likely to produce what you want the first time. labs.google/fx/tools/ima...
| 0 |
2025-01-22T21:14:40.131Z
|
2025-01-22T21:14:53.133Z
|
emollick.bsky.social
|
did:plc:flxq4uyjfotciovpw3x3fxnu
|
at://did:plc:flxq4uyjfotciovpw3x3fxnu/app.bsky.feed.post/3lgebp2eam22z
|
bafyreie47xlodvhzt5ffux7jpgaaqezbhxbqcdlnmkxqnucd4nl3d2ypta
|
<start>When people ask about the image generator they should use, I point them at Google’s ImageFX for Imagen 3.
Midjourney is better for art, Recraft is better for workflows, etc. but Imagen 3 is just the easiest to use and most likely to produce what you want the first time. labs.google/fx/tools/ima...
@emollick.bsky.social
|
When people ask about the image generator they should use, I point them at Google’s ImageFX for Imagen 3.
Midjourney is better for art, Recraft is better for workflows, etc. but Imagen 3 is just the easiest to use and most likely to produce what you want the first time. labs.google/fx/tools/ima...
| 0 |
2025-01-22T21:09:50.432Z
|
2025-01-22T21:10:07.226Z
|
emollick.bsky.social
|
did:plc:flxq4uyjfotciovpw3x3fxnu
|
at://did:plc:flxq4uyjfotciovpw3x3fxnu/app.bsky.feed.post/3lgdxl4oqec2z
|
bafyreibzwk7iptuxhf777avszkuuvizxgdl2yr2hjkxyojbofagajx2hiu
|
<start>This is Hailou’s S2V-01. I know other companies are releasing competitors soon. The workflow is literally just the single picture and the text prompt I show in the video. Shakespeare was generated on the first try, the mech was third try.
@emollick.bsky.social
|
This is Hailou’s S2V-01. I know other companies are releasing competitors soon. The workflow is literally just the single picture and the text prompt I show in the video. Shakespeare was generated on the first try, the mech was third try.
| 0 |
2025-01-22T18:08:41.242Z
|
2025-01-22T18:08:41.522Z
|
emollick.bsky.social
|
did:plc:flxq4uyjfotciovpw3x3fxnu
|
at://did:plc:flxq4uyjfotciovpw3x3fxnu/app.bsky.feed.post/3lgdvcmqyws26
|
bafyreigmwb5uewfeyqo6tukwb4vttw6hno3isppivxhdriiplkvfdl4gpu
|
<start>The new ability of AI video creators to add real people and products to scenes with just an image is likely to increase the utility (& more worryingly, misuse) of AI video.
Here I made Shakespeare at a cafe and the Girl with the Pearl Earring piloting a mech (just as Vermeer intended) This is Hailou’s S2V-01. I know other companies are releasing competitors soon. The workflow is literally just the single picture and the text prompt I show in the video. Shakespeare was generated on the first try, the mech was third try.
@emollick.bsky.social
|
The new ability of AI video creators to add real people and products to scenes with just an image is likely to increase the utility (& more worryingly, misuse) of AI video.
Here I made Shakespeare at a cafe and the Girl with the Pearl Earring piloting a mech (just as Vermeer intended)
| 1 |
2025-01-22T17:28:08.624Z
|
2025-01-22T17:28:08.733Z
|
emollick.bsky.social
|
did:plc:flxq4uyjfotciovpw3x3fxnu
|
at://did:plc:flxq4uyjfotciovpw3x3fxnu/app.bsky.feed.post/3lgdgksf4jc2t
|
bafyreicpndjzqt7pltgcmmhywqbplppg7mttt6fc5g3lo7uif6wqqfipka
|
<start>Attempt at a serious back-of-the-envelope calculation by Rohit Krishnan for how many workers a theoretical AGI would be equivalent to, given what we know about power and chip constraints - about 131M by 2030 in the base case, but there are lots of unknowns. www.strangeloopcanon.com/p/what-would...
@emollick.bsky.social
|
Attempt at a serious back-of-the-envelope calculation by Rohit Krishnan for how many workers a theoretical AGI would be equivalent to, given what we know about power and chip constraints - about 131M by 2030 in the base case, but there are lots of unknowns. www.strangeloopcanon.com/p/what-would...
| 0 |
2025-01-22T13:04:16.831Z
|
2025-01-22T13:04:21.117Z
|
emollick.bsky.social
|
did:plc:flxq4uyjfotciovpw3x3fxnu
|
at://did:plc:flxq4uyjfotciovpw3x3fxnu/app.bsky.feed.post/3lgcgongwnc2o
|
bafyreigqnv5b5prc3zm3goqlwdv5nnu234kbfnjqg7vokiteu57wesb5ny
|
<start>The only science fiction author who got AI right was Douglas Adams.
@emollick.bsky.social
|
The only science fiction author who got AI right was Douglas Adams.
| 0 |
2025-01-22T03:33:46.128Z
|
2025-01-22T03:33:47.230Z
|
emollick.bsky.social
|
did:plc:flxq4uyjfotciovpw3x3fxnu
|
at://did:plc:flxq4uyjfotciovpw3x3fxnu/app.bsky.feed.post/3lgcfyryiec2o
|
bafyreicxuvnpiievlxo7jk7553zhgvxp5akzoggmmkp44aqfxmkyhpqmpm
|
<start>No matter how much you fight it, I find that the visible chain-of-thought from DeepSeek makes it nearly impossible to avoid anthropomorphizing the thing
The visible first-person "thinking" makes you feel like you are reading a diary of a somewhat tortured soul who wants to help. Weird interaction The only science fiction author who got AI right was Douglas Adams.
@emollick.bsky.social
|
No matter how much you fight it, I find that the visible chain-of-thought from DeepSeek makes it nearly impossible to avoid anthropomorphizing the thing
The visible first-person "thinking" makes you feel like you are reading a diary of a somewhat tortured soul who wants to help. Weird interaction
| 1 |
2025-01-22T03:21:32.699Z
|
2025-01-22T03:21:34.117Z
|
emollick.bsky.social
|
did:plc:flxq4uyjfotciovpw3x3fxnu
|
at://did:plc:flxq4uyjfotciovpw3x3fxnu/app.bsky.feed.post/3lgcbur7lj22m
|
bafyreieh7jhbudgjrm222k7r7zzpi7bcomfjw6mgxxd3frc4xqihh5cicq
|
<start>These are very large gains in ability for a small model over the course of a month.
@emollick.bsky.social
|
These are very large gains in ability for a small model over the course of a month.
| 0 |
2025-01-22T02:07:42.697Z
|
2025-01-22T02:07:42.921Z
|
emollick.bsky.social
|
did:plc:flxq4uyjfotciovpw3x3fxnu
|
at://did:plc:flxq4uyjfotciovpw3x3fxnu/app.bsky.feed.post/3lgbct6wa222f
|
bafyreihwvf7rk3qbchpw57j52lauupgifoo2lza5htxlxgn4vgiiekgzdq
|
<start>This whole approach is still in the early days, but it has come far in less than a year: vvictoryuki.github.io/gamefactory/
@emollick.bsky.social
|
This whole approach is still in the early days, but it has come far in less than a year: vvictoryuki.github.io/gamefactory/
| 0 |
2025-01-21T16:52:03.964Z
|
2025-01-21T16:52:05.627Z
|
emollick.bsky.social
|
did:plc:flxq4uyjfotciovpw3x3fxnu
|
at://did:plc:flxq4uyjfotciovpw3x3fxnu/app.bsky.feed.post/3lgbcpiqy222f
|
bafyreigryusjznrmhbcyvmh2miedjqacpazy5gcbuzszax3bnd5tywdzcy
|
<start>This seems like the next step in the rapid development towards dynamic playable games generated on-the-fly by AI.
This research group trained a world model on Minecraft and then put a diffusion model on top, so that they can create a novel real-time 3D game set in any location they prompt This whole approach is still in the early days, but it has come far in less than a year: vvictoryuki.github.io/gamefactory/
@emollick.bsky.social
|
This seems like the next step in the rapid development towards dynamic playable games generated on-the-fly by AI.
This research group trained a world model on Minecraft and then put a diffusion model on top, so that they can create a novel real-time 3D game set in any location they prompt
| 1 |
2025-01-21T16:50:00.065Z
|
2025-01-21T16:50:01.030Z
|
emollick.bsky.social
|
did:plc:flxq4uyjfotciovpw3x3fxnu
|
at://did:plc:flxq4uyjfotciovpw3x3fxnu/app.bsky.feed.post/3lg7wygklhs2e
|
bafyreie7pi6i4xdumaguec5gs33zbu2hjz3hvpjh5gutby5ljr7c5sx3ua
|
<start>If it "knows" and "likes" your work, you likely have a better chance of being suggested to future scholars. Its weird, but here we are.
@emollick.bsky.social
|
If it "knows" and "likes" your work, you likely have a better chance of being suggested to future scholars. Its weird, but here we are.
| 0 |
2025-01-21T03:47:35.108Z
|
2025-01-21T03:47:35.620Z
|
emollick.bsky.social
|
did:plc:flxq4uyjfotciovpw3x3fxnu
|
at://did:plc:flxq4uyjfotciovpw3x3fxnu/app.bsky.feed.post/3lg7wc2ain22m
|
bafyreidcdexfzkck7clawhu4lv4ub5bx6e7sh63oxczoay44cbuygbgfku
|
<start>I think academics may need to start writing for AI
Here is a bit of the internal monologue of DeepSeek R1 when I asked it come up with a sociological theory (the details don’t matter). It came up with a name and then realized it was already a name for a theory proposed by March, so it built on that If it "knows" and "likes" your work, you likely have a better chance of being suggested to future scholars. Its weird, but here we are.
@emollick.bsky.social
|
I think academics may need to start writing for AI
Here is a bit of the internal monologue of DeepSeek R1 when I asked it come up with a sociological theory (the details don’t matter). It came up with a name and then realized it was already a name for a theory proposed by March, so it built on that
| 1 |
2025-01-21T03:35:03.990Z
|
2025-01-21T03:35:05.427Z
|
emollick.bsky.social
|
did:plc:flxq4uyjfotciovpw3x3fxnu
|
at://did:plc:flxq4uyjfotciovpw3x3fxnu/app.bsky.feed.post/3lg7tc43hss2m
|
bafyreicaqptdanx6fsrxgwg5m5vwaxm5ld65cvnfq6h3aqkvir4gewgzr4
|
<start>Apparently reasoning LLMs can apply theory-of-mind in their reasoning (and are pretty solid at sociology): “Think like a cat raised among birds and come up with a new form of neo-institutional legitimation”
@emollick.bsky.social
|
Apparently reasoning LLMs can apply theory-of-mind in their reasoning (and are pretty solid at sociology): “Think like a cat raised among birds and come up with a new form of neo-institutional legitimation”
| 0 |
2025-01-21T02:41:24.697Z
|
2025-01-21T02:41:26.333Z
|
emollick.bsky.social
|
did:plc:flxq4uyjfotciovpw3x3fxnu
|
at://did:plc:flxq4uyjfotciovpw3x3fxnu/app.bsky.feed.post/3lg7cedevok2e
|
bafyreig6ypppwhcrgrs3qrijcsqazi4xjumjtyxanbac4vj2oy4gjrrfva
|
<start>Surprisingly (at least to me) Claude produced a worse verse, which basically reads as a digital parody of the original poem.
@emollick.bsky.social
|
Surprisingly (at least to me) Claude produced a worse verse, which basically reads as a digital parody of the original poem.
| 0 |
2025-01-20T21:38:25.847Z
|
2025-01-20T21:38:36.421Z
|
emollick.bsky.social
|
did:plc:flxq4uyjfotciovpw3x3fxnu
|
at://did:plc:flxq4uyjfotciovpw3x3fxnu/app.bsky.feed.post/3lg7bvazj2s2e
|
bafyreihsxql36ig5c75ydpbamvml45l5nrplturlzpzj5kw4ts4vuibv5u
|
<start>Asking DeepSeek to add a section to The Wasteland is interesting less for the completed verse (which has a few *great* images but is otherwise too random) than the fact that the chain of thought produced by the AI is essentially literary criticism & analysis of Eliot’s work. Surprisingly (at least to me) Claude produced a worse verse, which basically reads as a digital parody of the original poem.
@emollick.bsky.social
|
Asking DeepSeek to add a section to The Wasteland is interesting less for the completed verse (which has a few *great* images but is otherwise too random) than the fact that the chain of thought produced by the AI is essentially literary criticism & analysis of Eliot’s work.
| 1 |
2025-01-20T21:30:00.057Z
|
2025-01-20T21:30:14.231Z
|
emollick.bsky.social
|
did:plc:flxq4uyjfotciovpw3x3fxnu
|
at://did:plc:flxq4uyjfotciovpw3x3fxnu/app.bsky.feed.post/3lg73y6ekx22r
|
bafyreigiao36kpiv726axsltjbbpsougzlaq4gn5zjqyinc37sf4phycaa
|
<start>There’s a subtle psychological impact also. As opposed to models that just answer you, or even o1’s distilled version, the chain of thought with its little bits of hesitation & near anxiety makes you really feel for the model: “The poor little LLM is working so hard for me” Apparently reasoning LLMs can apply theory-of-mind in their reasoning (and are pretty solid at sociology): “Think like a cat raised among birds and come up with a new form of neo-institutional legitimation”
@emollick.bsky.social
|
There’s a subtle psychological impact also. As opposed to models that just answer you, or even o1’s distilled version, the chain of thought with its little bits of hesitation & near anxiety makes you really feel for the model: “The poor little LLM is working so hard for me”
| 1 |
2025-01-20T19:44:15.490Z
|
2025-01-20T19:44:19.617Z
|
emollick.bsky.social
|
did:plc:flxq4uyjfotciovpw3x3fxnu
|
at://did:plc:flxq4uyjfotciovpw3x3fxnu/app.bsky.feed.post/3lg72yjl7dk2r
|
bafyreiau3tfsfyix3lbdliegy2jhhf6onpigjychk2f3qs75cjjqe2nz2y
|
<start>The raw chain of thought from DeepSeek is fascinating, really reads like a human thinking out loud. Charming and strange. There’s a subtle psychological impact also. As opposed to models that just answer you, or even o1’s distilled version, the chain of thought with its little bits of hesitation & near anxiety makes you really feel for the model: “The poor little LLM is working so hard for me”
@emollick.bsky.social
|
The raw chain of thought from DeepSeek is fascinating, really reads like a human thinking out loud. Charming and strange.
| 1 |
2025-01-20T19:26:33.494Z
|
2025-01-20T19:26:35.821Z
|
emollick.bsky.social
|
did:plc:flxq4uyjfotciovpw3x3fxnu
|
at://did:plc:flxq4uyjfotciovpw3x3fxnu/app.bsky.feed.post/3lg6x74jjfc2r
|
bafyreibhm5ysvxmzyjmoi6xvf7mxzkwrfnlnjjgjoi2qxnjisol6ttyqn4
|
<start>I should clarify that you likely cannot restrict access to frontier models, but that doesn’t mean that all regulation is useless.
@emollick.bsky.social
|
I should clarify that you likely cannot restrict access to frontier models, but that doesn’t mean that all regulation is useless.
| 0 |
2025-01-20T18:18:39.730Z
|
2025-01-20T18:18:40.218Z
|
emollick.bsky.social
|
did:plc:flxq4uyjfotciovpw3x3fxnu
|
at://did:plc:flxq4uyjfotciovpw3x3fxnu/app.bsky.feed.post/3lg6uasnipc2r
|
bafyreica5qwaaw44qhpae7xpona2qaaaxs4sayyblez7cayw7rvsulkgky
|
<start>The release of DeepSeek-R1 demonstrates that, for better or worse, any attempt to restrict AI by national governments is unlikely to work in the future. You can get an open frontier model on a USB stick, and the methods outlined by DeepSeek suggest pathways forward for other open models, too. I should clarify that you likely cannot restrict access to frontier models, but that doesn’t mean that all regulation is useless.
@emollick.bsky.social
|
The release of DeepSeek-R1 demonstrates that, for better or worse, any attempt to restrict AI by national governments is unlikely to work in the future. You can get an open frontier model on a USB stick, and the methods outlined by DeepSeek suggest pathways forward for other open models, too.
| 1 |
2025-01-20T17:25:55.254Z
|
2025-01-20T17:25:55.633Z
|
emollick.bsky.social
|
did:plc:flxq4uyjfotciovpw3x3fxnu
|
at://did:plc:flxq4uyjfotciovpw3x3fxnu/app.bsky.feed.post/3lg6ofln6r22a
|
bafyreigqkz35lnakjc4v5lyfvrn3ddlul7fopy342nxy5rgouiq5hqzs5m
|
<start>I do like where it rationalizes its own laziness in not creating quite the poem I asked for.
@emollick.bsky.social
|
I do like where it rationalizes its own laziness in not creating quite the poem I asked for.
| 0 |
2025-01-20T15:41:13.229Z
|
2025-01-20T15:41:14.424Z
|
emollick.bsky.social
|
did:plc:flxq4uyjfotciovpw3x3fxnu
|
at://did:plc:flxq4uyjfotciovpw3x3fxnu/app.bsky.feed.post/3lg6mvspers2a
|
bafyreieyrskusygjvu53pfnwmvkjlp2exsjznur27trb34ozheurkeazlm
|
<start>The smallest versions can run on my laptop. This is a pretty big leap, because the only reasoning (test-time compute) models so far have been from Google and OpenAI. You can try it: chat.deepseek.com I do like where it rationalizes its own laziness in not creating quite the poem I asked for.
@emollick.bsky.social
|
The smallest versions can run on my laptop. This is a pretty big leap, because the only reasoning (test-time compute) models so far have been from Google and OpenAI. You can try it: chat.deepseek.com
| 1 |
2025-01-20T15:14:30.028Z
|
2025-01-20T15:14:31.218Z
|
emollick.bsky.social
|
did:plc:flxq4uyjfotciovpw3x3fxnu
|
at://did:plc:flxq4uyjfotciovpw3x3fxnu/app.bsky.feed.post/3lg6ml4632s2a
|
bafyreicapvcs5schdas4aik3cs7de56ds7m7toeiojh4ilhguyl3ccnnyu
|
<start>A new completely open reasoning model out of China, Deepseek-R1, is now available. The benchmarks show it at parity with the likes of o1 and Sonnet
In some informal tests on non-code problems, it is really good, not o1-pro level but surprisingly capable (and incredibly small & fast!). Big advance. The smallest versions can run on my laptop. This is a pretty big leap, because the only reasoning (test-time compute) models so far have been from Google and OpenAI. You can try it: chat.deepseek.com
@emollick.bsky.social
|
A new completely open reasoning model out of China, Deepseek-R1, is now available. The benchmarks show it at parity with the likes of o1 and Sonnet
In some informal tests on non-code problems, it is really good, not o1-pro level but surprisingly capable (and incredibly small & fast!). Big advance.
| 1 |
2025-01-20T15:08:30.836Z
|
2025-01-20T15:08:34.320Z
|
emollick.bsky.social
|
did:plc:flxq4uyjfotciovpw3x3fxnu
|
at://did:plc:flxq4uyjfotciovpw3x3fxnu/app.bsky.feed.post/3lg5qwfoeok2a
|
bafyreicrfogix4u4kw2gekueu6zu42sqjdg3b4iplyoec3tyesdsgpk2ze
|
<start>This video doesn't communicate much (other than the scene I described), but I have created videos that have a message and meaning, at least to me. I am not sure what that means - is what you can create with AI art? terrible pastiche? something else entirely? I don't have answers, maybe you do?
@emollick.bsky.social
|
This video doesn't communicate much (other than the scene I described), but I have created videos that have a message and meaning, at least to me. I am not sure what that means - is what you can create with AI art? terrible pastiche? something else entirely? I don't have answers, maybe you do?
| 0 |
2025-01-20T06:53:45.147Z
|
2025-01-20T06:53:43.915Z
|
emollick.bsky.social
|
did:plc:flxq4uyjfotciovpw3x3fxnu
|
at://did:plc:flxq4uyjfotciovpw3x3fxnu/app.bsky.feed.post/3lg5qgbu6gs2a
|
bafyreiedtb2pcvous6kmelluyzrikphbac24le2q5oeq6a3jprod3uidpy
|
<start>There is still a very large gap between the control that film gives to creators and the strangeness of using AI video generators which involves more randomness. The gap may be closed eventually (soon?) but for now you can create some interesting, if incoherent, videos easily.
@emollick.bsky.social
|
There is still a very large gap between the control that film gives to creators and the strangeness of using AI video generators which involves more randomness. The gap may be closed eventually (soon?) but for now you can create some interesting, if incoherent, videos easily.
| 0 |
2025-01-20T06:44:44.272Z
|
2025-01-20T06:44:43.016Z
|
emollick.bsky.social
|
did:plc:flxq4uyjfotciovpw3x3fxnu
|
at://did:plc:flxq4uyjfotciovpw3x3fxnu/app.bsky.feed.post/3lg5qb42wuc2a
|
bafyreifisiaj4tjxxmurjh43g33imkuvpabxw3zn5mxtddqo2654i4lqp4
|
<start>How does veo 2 pull off three nervous women holding knives on the back of a giant caterpillar running through a deserted city looking determined as a tank rolls towards them?
If I imagine a thing, I can generate something close. It isn't a replacement for film's coherence, it is a new kind of thing There is still a very large gap between the control that film gives to creators and the strangeness of using AI video generators which involves more randomness. The gap may be closed eventually (soon?) but for now you can create some interesting, if incoherent, videos easily.
@emollick.bsky.social
|
How does veo 2 pull off three nervous women holding knives on the back of a giant caterpillar running through a deserted city looking determined as a tank rolls towards them?
If I imagine a thing, I can generate something close. It isn't a replacement for film's coherence, it is a new kind of thing
| 1 |
2025-01-20T06:41:50.430Z
|
2025-01-20T06:41:49.718Z
|
emollick.bsky.social
|
did:plc:flxq4uyjfotciovpw3x3fxnu
|
at://did:plc:flxq4uyjfotciovpw3x3fxnu/app.bsky.feed.post/3lg4hnlt2ck2d
|
bafyreid63k54yax4en2jhz2yuouhbtbl2lpcw5uf7uvaphxqqs3bnoo5ni
|
<start>There are no rigorous studies yet on how to best prompt reasoning models like o1.
They are not chat models, they are designed so that you give them a question and you get an answer. You want to give them context, clarity, etc. but we don’t know what the optimal way to do that is yet.
@emollick.bsky.social
|
There are no rigorous studies yet on how to best prompt reasoning models like o1.
They are not chat models, they are designed so that you give them a question and you get an answer. You want to give them context, clarity, etc. but we don’t know what the optimal way to do that is yet.
| 0 |
2025-01-19T18:35:06.186Z
|
2025-01-19T18:35:06.423Z
|
emollick.bsky.social
|
did:plc:flxq4uyjfotciovpw3x3fxnu
|
at://did:plc:flxq4uyjfotciovpw3x3fxnu/app.bsky.feed.post/3lg4darqwfc2d
|
bafyreigd4db2iyck6c3okcquzdzgsuigqelyuvegr7ah5no57qgquxdh3q
|
<start>AIs in the world can create complex systems with unexpected & risky feedback loops.
Each output affects the world, affecting inputs, affecting outputs... Even simple loops can drive optimization behavior, making systems turn extreme. No training needed, just interaction. arxiv.org/pdf/2402.06627
@emollick.bsky.social
|
AIs in the world can create complex systems with unexpected & risky feedback loops.
Each output affects the world, affecting inputs, affecting outputs... Even simple loops can drive optimization behavior, making systems turn extreme. No training needed, just interaction. arxiv.org/pdf/2402.06627
| 0 |
2025-01-19T17:16:21.226Z
|
2025-01-19T17:16:23.728Z
|
emollick.bsky.social
|
did:plc:flxq4uyjfotciovpw3x3fxnu
|
at://did:plc:flxq4uyjfotciovpw3x3fxnu/app.bsky.feed.post/3lg25tclpj22y
|
bafyreiaxtfbbhu6w6chhz3h3woiafervlh4bqplv2tudjwbztksramk6cm
|
<start>Yet more evidence that people can’t accurately detect well-prompted AI writing (and AI can’t accurately detect well-prompted AI writing, either). arxiv.org/pdf/2407.08853
@emollick.bsky.social
|
Yet more evidence that people can’t accurately detect well-prompted AI writing (and AI can’t accurately detect well-prompted AI writing, either). arxiv.org/pdf/2407.08853
| 0 |
2025-01-18T20:34:00.937Z
|
2025-01-18T20:34:09.829Z
|
emollick.bsky.social
|
did:plc:flxq4uyjfotciovpw3x3fxnu
|
at://did:plc:flxq4uyjfotciovpw3x3fxnu/app.bsky.feed.post/3lfzxlqtexc2s
|
bafyreidj7uea3i2ubifivzbh2uffos6rga4kc6d5et5xldjhndzo6idndm
|
<start>It would help conversations to specify where the hype is and where the value is.
@emollick.bsky.social
|
It would help conversations to specify where the hype is and where the value is.
| 0 |
2025-01-18T18:42:24.986Z
|
2025-01-18T18:42:27.224Z
|
emollick.bsky.social
|
did:plc:flxq4uyjfotciovpw3x3fxnu
|
at://did:plc:flxq4uyjfotciovpw3x3fxnu/app.bsky.feed.post/3lfzxiaswsk2s
|
bafyreih55c2lwexbpc2hrlzin6zwflps6w4g34gxfr4flf53geajiwgphy
|
<start>And we now have more than enough real world evidence from controlled trials (and tons more case studies etc) to suggest that there is real value for real organizations in AI use in some (not all) cases.
Combine that with the growing number of open models & this is not something that is going away.
@emollick.bsky.social
|
And we now have more than enough real world evidence from controlled trials (and tons more case studies etc) to suggest that there is real value for real organizations in AI use in some (not all) cases.
Combine that with the growing number of open models & this is not something that is going away.
| 0 |
2025-01-18T18:40:27.531Z
|
2025-01-18T18:40:29.523Z
|
emollick.bsky.social
|
did:plc:flxq4uyjfotciovpw3x3fxnu
|
at://did:plc:flxq4uyjfotciovpw3x3fxnu/app.bsky.feed.post/3lfzxfxixjc2s
|
bafyreibtvbmv2q3mxnwjyzp5ouuhoencseaxu2uzn7nvh6qveuaewktcw4
|
<start>A surprising percentage of people I speak to think that AI is just going to disappear because of these arguments, and aren’t thinking about how to address AI as a result
To the extent there is a “AI bubble” financially (and I am not convinced there is) that doesn’t mean AI goes away if it pops And we now have more than enough real world evidence from controlled trials (and tons more case studies etc) to suggest that there is real value for real organizations in AI use in some (not all) cases.
Combine that with the growing number of open models & this is not something that is going away.
@emollick.bsky.social
|
A surprising percentage of people I speak to think that AI is just going to disappear because of these arguments, and aren’t thinking about how to address AI as a result
To the extent there is a “AI bubble” financially (and I am not convinced there is) that doesn’t mean AI goes away if it pops
| 1 |
2025-01-18T18:39:10.658Z
|
2025-01-18T18:39:11.820Z
|
emollick.bsky.social
|
did:plc:flxq4uyjfotciovpw3x3fxnu
|
at://did:plc:flxq4uyjfotciovpw3x3fxnu/app.bsky.feed.post/3lfzx5lbask2s
|
bafyreif75mk234soylx5anr2eqogjkbhoh4pzltum3c7fphkshb33rhqxm
|
<start>It is good that there are critics of AI, there are lots of things to critique.
But I think it is a disservice to suggest that AI is going to “go away” because it is useless/due to model collapse/lawsuits/etc. Not true & it prevents us from facing the actual challenges & getting potential benefits. A surprising percentage of people I speak to think that AI is just going to disappear because of these arguments, and aren’t thinking about how to address AI as a result
To the extent there is a “AI bubble” financially (and I am not convinced there is) that doesn’t mean AI goes away if it pops
@emollick.bsky.social
|
It is good that there are critics of AI, there are lots of things to critique.
But I think it is a disservice to suggest that AI is going to “go away” because it is useless/due to model collapse/lawsuits/etc. Not true & it prevents us from facing the actual challenges & getting potential benefits.
| 1 |
2025-01-18T18:34:29.387Z
|
2025-01-18T18:34:30.524Z
|
emollick.bsky.social
|
did:plc:flxq4uyjfotciovpw3x3fxnu
|
at://did:plc:flxq4uyjfotciovpw3x3fxnu/app.bsky.feed.post/3lfzwxczsps2s
|
bafyreiga4qqxe3pejzdsfsbmd4lnwrz6kwcrh4lmg23pi2xuz6avjswt7y
|
<start>Useful thing for AI critics to do would is to "red team" their own beliefs by spending some time to see if they can get AI to do the things they say it can't
This isn't intended to be snarky, there are lots of things AI does terribly. But we need more critics to actually engage with these systems It is good that there are critics of AI, there are lots of things to critique.
But I think it is a disservice to suggest that AI is going to “go away” because it is useless/due to model collapse/lawsuits/etc. Not true & it prevents us from facing the actual challenges & getting potential benefits.
@emollick.bsky.social
|
Useful thing for AI critics to do would is to "red team" their own beliefs by spending some time to see if they can get AI to do the things they say it can't
This isn't intended to be snarky, there are lots of things AI does terribly. But we need more critics to actually engage with these systems
| 1 |
2025-01-18T18:30:59.426Z
|
2025-01-18T18:31:00.722Z
|
emollick.bsky.social
|
did:plc:flxq4uyjfotciovpw3x3fxnu
|
at://did:plc:flxq4uyjfotciovpw3x3fxnu/app.bsky.feed.post/3lfzezobz2k26
|
bafyreihkkhob6rkbdtbhd4zkjtzibzoatu45wz7l62smbktqbiawtmoz4y
|
<start>AIs beat humans in basically all creativity tests like the Torrence Test, but it isn't clear what exactly that means compared to humans.
@emollick.bsky.social
|
AIs beat humans in basically all creativity tests like the Torrence Test, but it isn't clear what exactly that means compared to humans.
| 0 |
2025-01-18T13:10:10.986Z
|
2025-01-18T13:10:12.317Z
|
emollick.bsky.social
|
did:plc:flxq4uyjfotciovpw3x3fxnu
|
at://did:plc:flxq4uyjfotciovpw3x3fxnu/app.bsky.feed.post/3lfzed3axcs26
|
bafyreibbfbz6g4yvrebgwbtvswcxpsdt64j7kz3eczwwm6wto3laeuraju
|
<start>To make this less cryptic for those who don't know. We have a bunch of tests that are predictive of things like social performance on teams, and so can be used to measure "social intelligence." But it is unclear if they would measure anything useful for AI. socialintelligence.labinthewild.org/mite/ AIs beat humans in basically all creativity tests like the Torrence Test, but it isn't clear what exactly that means compared to humans.
@emollick.bsky.social
|
To make this less cryptic for those who don't know. We have a bunch of tests that are predictive of things like social performance on teams, and so can be used to measure "social intelligence." But it is unclear if they would measure anything useful for AI. socialintelligence.labinthewild.org/mite/
| 1 |
2025-01-18T12:57:32.830Z
|
2025-01-18T12:57:34.022Z
|
emollick.bsky.social
|
did:plc:flxq4uyjfotciovpw3x3fxnu
|
at://did:plc:flxq4uyjfotciovpw3x3fxnu/app.bsky.feed.post/3lfze764kos26
|
bafyreihtuhkje22szs5xpovkrzw7mpess6kks7opwjrf45trxjqeqehari
|
<start>Our lack of good deep psychological measures of human creativity, reasoning, empathy, etc. is really a problem in AI right now.
A lot of tests that were "good enough" for human research (RAT for creativity, Seeing the Mind in The Eyes for empathy) are not robust enough to be benchmarks for AI. To make this less cryptic for those who don't know. We have a bunch of tests that are predictive of things like social performance on teams, and so can be used to measure "social intelligence." But it is unclear if they would measure anything useful for AI. socialintelligence.labinthewild.org/mite/
@emollick.bsky.social
|
Our lack of good deep psychological measures of human creativity, reasoning, empathy, etc. is really a problem in AI right now.
A lot of tests that were "good enough" for human research (RAT for creativity, Seeing the Mind in The Eyes for empathy) are not robust enough to be benchmarks for AI.
| 1 |
2025-01-18T12:55:21.616Z
|
2025-01-18T12:55:21.919Z
|
emollick.bsky.social
|
did:plc:flxq4uyjfotciovpw3x3fxnu
|
at://did:plc:flxq4uyjfotciovpw3x3fxnu/app.bsky.feed.post/3lfymqgdgak2e
|
bafyreievgqk7mzetcvwkruwzbk4cchuodi4isabprbiazshjjmd5uzviby
|
<start>I wrote about my experience with Claude as an agent, for example: www.oneusefulthing.org/p/when-you-g...
@emollick.bsky.social
|
I wrote about my experience with Claude as an agent, for example: www.oneusefulthing.org/p/when-you-g...
| 0 |
2025-01-18T05:55:30.851Z
|
2025-01-18T05:55:31.921Z
|
emollick.bsky.social
|
did:plc:flxq4uyjfotciovpw3x3fxnu
|
at://did:plc:flxq4uyjfotciovpw3x3fxnu/app.bsky.feed.post/3lfympnmbtk2e
|
bafyreiaxdatfjqzkmw2grs7d3vrppsmntlrki4n4e6pmp27upknxqxxk6a
|
<start>An agent can handle some very complicated tasks if it is in a narrow domain with good prompting and tools, but, interestingly, any time building narrow agents will feel like a waste if better models come along and solve the general agent use case, which is also possible. I wrote about my experience with Claude as an agent, for example: www.oneusefulthing.org/p/when-you-g...
@emollick.bsky.social
|
An agent can handle some very complicated tasks if it is in a narrow domain with good prompting and tools, but, interestingly, any time building narrow agents will feel like a waste if better models come along and solve the general agent use case, which is also possible.
| 1 |
2025-01-18T05:55:04.927Z
|
2025-01-18T05:55:05.221Z
|
emollick.bsky.social
|
did:plc:flxq4uyjfotciovpw3x3fxnu
|
at://did:plc:flxq4uyjfotciovpw3x3fxnu/app.bsky.feed.post/3lfymommves2e
|
bafyreiaovehgfrstv3szy2pnqryilquwf2z6cwgn5m6jteq6a4uoj5tsp4
|
<start>I have spent a lot of time with AI agents (including Devin and Claude Computer Use) and they really do remain too fragile & not "smart" enough to be reliable for complicated tasks.
So you have two options: (1) wait for better models or (2) focus on narrower use cases (like Gemini Deep Research). An agent can handle some very complicated tasks if it is in a narrow domain with good prompting and tools, but, interestingly, any time building narrow agents will feel like a waste if better models come along and solve the general agent use case, which is also possible.
@emollick.bsky.social
|
I have spent a lot of time with AI agents (including Devin and Claude Computer Use) and they really do remain too fragile & not "smart" enough to be reliable for complicated tasks.
So you have two options: (1) wait for better models or (2) focus on narrower use cases (like Gemini Deep Research).
| 1 |
2025-01-18T05:54:30.344Z
|
2025-01-18T05:54:30.719Z
|
emollick.bsky.social
|
did:plc:flxq4uyjfotciovpw3x3fxnu
|
at://did:plc:flxq4uyjfotciovpw3x3fxnu/app.bsky.feed.post/3lfxn3xih7c2c
|
bafyreidmpfk3ch2jjh3tiogz7u5zup4ii5i6fjlrzmrqstcet52makutmi
|
<start>I have been working with all the video tools for awhile and post many examples (video doesn't work well on BlueSky right now, making it hard to share here).
I am not cherrypicking in these cases, but these systems are also still very flawed. But there is a rapid improvement in progress & utility.
@emollick.bsky.social
|
I have been working with all the video tools for awhile and post many examples (video doesn't work well on BlueSky right now, making it hard to share here).
I am not cherrypicking in these cases, but these systems are also still very flawed. But there is a rapid improvement in progress & utility.
| 0 |
2025-01-17T20:29:18.201Z
|
2025-01-17T20:29:18.516Z
|
emollick.bsky.social
|
did:plc:flxq4uyjfotciovpw3x3fxnu
|
at://did:plc:flxq4uyjfotciovpw3x3fxnu/app.bsky.feed.post/3lfxm6j33322c
|
bafyreihirtg5klzjtwngtbtjnzfjlyqgn7pw7uisd2uhvl7z7oy7w33mai
|
<start>It is odd that the world's leading AI lab, producing a system that they consider pivotal to the future and also potentially dangerous, communicates their product development progress primarily through vague and oracular X posts. Its entertaining, but also really weird.
@emollick.bsky.social
|
It is odd that the world's leading AI lab, producing a system that they consider pivotal to the future and also potentially dangerous, communicates their product development progress primarily through vague and oracular X posts. Its entertaining, but also really weird.
| 0 |
2025-01-17T20:12:50.003Z
|
2025-01-17T20:12:51.219Z
|
emollick.bsky.social
|
did:plc:flxq4uyjfotciovpw3x3fxnu
|
at://did:plc:flxq4uyjfotciovpw3x3fxnu/app.bsky.feed.post/3lfxixagg222c
|
bafyreibnf5ylr6lejyl2c76a6u6gik7glxffekjekie73funmeo6mhny4y
|
<start>Most of the problems I have with veo 2 are when I don't describe something well. There is still a lot of randomness and flaws (especially if you pause or watch a second time) but we are getting close to an actually useful tool rather than a roulette wheel of a video clip generator.
@emollick.bsky.social
|
Most of the problems I have with veo 2 are when I don't describe something well. There is still a lot of randomness and flaws (especially if you pause or watch a second time) but we are getting close to an actually useful tool rather than a roulette wheel of a video clip generator.
| 0 |
2025-01-17T19:15:04.833Z
|
2025-01-17T19:15:05.126Z
|
emollick.bsky.social
|
did:plc:flxq4uyjfotciovpw3x3fxnu
|
at://did:plc:flxq4uyjfotciovpw3x3fxnu/app.bsky.feed.post/3lfxiq4jjxs2c
|
bafyreic52deekwvrbryncbzfl24l3mtbjfvowlshz5idww2pj3ng5jillq
|
<start>How does "a closeup of a woman's face, she is wearing silver glasses adorned with a peacock motif, the shot cuts & she is grabbing a piece of paper off a gilt table in huge art deco room and throwing it into a burning brazier" work for me as a veo prompt the first time around?
Closer to a real tool Most of the problems I have with veo 2 are when I don't describe something well. There is still a lot of randomness and flaws (especially if you pause or watch a second time) but we are getting close to an actually useful tool rather than a roulette wheel of a video clip generator.
@emollick.bsky.social
|
How does "a closeup of a woman's face, she is wearing silver glasses adorned with a peacock motif, the shot cuts & she is grabbing a piece of paper off a gilt table in huge art deco room and throwing it into a burning brazier" work for me as a veo prompt the first time around?
Closer to a real tool
| 1 |
2025-01-17T19:11:05.860Z
|
2025-01-17T19:11:06.522Z
|
emollick.bsky.social
|
did:plc:flxq4uyjfotciovpw3x3fxnu
|
at://did:plc:flxq4uyjfotciovpw3x3fxnu/app.bsky.feed.post/3lfvlkjw4oc2c
|
bafyreih4dtswmsuu6armltfd6z55b2slty54wwyvzcj43bfpb2sahfwcsm
|
<start>"Hey Gemini, Claude, o1 & Grok: i need something I could make if I was transported back in time to 1100 AD Europe that would help me become a local lord. No magic tricks. Some invention I could reasonably make using the resources and tools available, and exact instructions." Some folks have suggested penicillin. o1-pro thinks not.
@emollick.bsky.social
|
"Hey Gemini, Claude, o1 & Grok: i need something I could make if I was transported back in time to 1100 AD Europe that would help me become a local lord. No magic tricks. Some invention I could reasonably make using the resources and tools available, and exact instructions."
| 1 |
2025-01-17T00:56:20.318Z
|
2025-01-17T00:56:22.322Z
|
emollick.bsky.social
|
did:plc:flxq4uyjfotciovpw3x3fxnu
|
at://did:plc:flxq4uyjfotciovpw3x3fxnu/app.bsky.feed.post/3lft7tia6tc2e
|
bafyreifkq7mecfxsmkqns2inaqhjbd2ytd3ftxbyrlmblhatu7xpvzmffy
|
<start>I wrote more about when AI helps and hurts education in this earlier post. www.oneusefulthing.org/p/post-apoca...
@emollick.bsky.social
|
I wrote more about when AI helps and hurts education in this earlier post. www.oneusefulthing.org/p/post-apoca...
| 0 |
2025-01-16T02:21:16.170Z
|
2025-01-16T02:21:17.029Z
|
emollick.bsky.social
|
did:plc:flxq4uyjfotciovpw3x3fxnu
|
at://did:plc:flxq4uyjfotciovpw3x3fxnu/app.bsky.feed.post/3lfsupgijes26
|
bafyreihad7oilqm2pok4zawt3yfvdt3apwcuwztc377v3jww5p2u33sqku
|
<start>And since it isn't clear to everyone who doesn't read the blog post - the fact that this is teacher-led is likely pretty important. We know that independent student use of AI as a tutor can harm learning in some circumstances, because it gives the illusion of learning. I wrote more about when AI helps and hurts education in this earlier post. www.oneusefulthing.org/p/post-apoca...
@emollick.bsky.social
|
And since it isn't clear to everyone who doesn't read the blog post - the fact that this is teacher-led is likely pretty important. We know that independent student use of AI as a tutor can harm learning in some circumstances, because it gives the illusion of learning.
| 1 |
2025-01-15T23:02:08.967Z
|
2025-01-15T23:02:09.522Z
|
emollick.bsky.social
|
did:plc:flxq4uyjfotciovpw3x3fxnu
|
at://did:plc:flxq4uyjfotciovpw3x3fxnu/app.bsky.feed.post/3lfsnwvqv2k26
|
bafyreide5yhdgk2gjpeqouqzbmdfr4yp72htnirbaftbcygjkkvn6xv3zi
|
<start>They used Microsoft Copilot and teachers provided guidance and initial prompts. No working paper yet, but the results and experiment are written up here: blogs.worldbank.org/en/education... And since it isn't clear to everyone who doesn't read the blog post - the fact that this is teacher-led is likely pretty important. We know that independent student use of AI as a tutor can harm learning in some circumstances, because it gives the illusion of learning.
@emollick.bsky.social
|
They used Microsoft Copilot and teachers provided guidance and initial prompts. No working paper yet, but the results and experiment are written up here: blogs.worldbank.org/en/education...
| 1 |
2025-01-15T21:01:03.659Z
|
2025-01-15T21:01:05.642Z
|
emollick.bsky.social
|
did:plc:flxq4uyjfotciovpw3x3fxnu
|
at://did:plc:flxq4uyjfotciovpw3x3fxnu/app.bsky.feed.post/3lfsnssanxs2e
|
bafyreihgx3pf473j7uahbau665wygxjbi7vouxo2ezhwsopdxghbgutlya
|
<start>New randomized, controlled trial by the World Bank of students using GPT-4 as a tutor in Nigeria. Six weeks of after-school AI tutoring = 2 years of typical learning gains, outperforming 80% of other educational interventions.
And it helped all students, especially girls who were initially behind. They used Microsoft Copilot and teachers provided guidance and initial prompts. No working paper yet, but the results and experiment are written up here: blogs.worldbank.org/en/education...
@emollick.bsky.social
|
New randomized, controlled trial by the World Bank of students using GPT-4 as a tutor in Nigeria. Six weeks of after-school AI tutoring = 2 years of typical learning gains, outperforming 80% of other educational interventions.
And it helped all students, especially girls who were initially behind.
| 1 |
2025-01-15T20:58:45.762Z
|
2025-01-15T20:58:47.621Z
|
emollick.bsky.social
|
did:plc:flxq4uyjfotciovpw3x3fxnu
|
at://did:plc:flxq4uyjfotciovpw3x3fxnu/app.bsky.feed.post/3lfsklqxcac26
|
bafyreien26yuhl5nztq4s4udqbebqmep3zpdcjmz4jxbjcqyfdikrxyovm
|
<start>Insane that this works in veo 2: "a fireproof woman in a wedding dress wades through knee-deep red burning lava, tugging at her hems as the dress burns, around her is a barren plain, in the distance is a brutalist building rising from a sea of spindly trees"
Look at the lighting, for example...
@emollick.bsky.social
|
Insane that this works in veo 2: "a fireproof woman in a wedding dress wades through knee-deep red burning lava, tugging at her hems as the dress burns, around her is a barren plain, in the distance is a brutalist building rising from a sea of spindly trees"
Look at the lighting, for example...
| 0 |
2025-01-15T20:01:08.302Z
|
2025-01-15T20:01:09.219Z
|
emollick.bsky.social
|
did:plc:flxq4uyjfotciovpw3x3fxnu
|
at://did:plc:flxq4uyjfotciovpw3x3fxnu/app.bsky.feed.post/3lfsan5q5zs26
|
bafyreih3sdidcitbyxzeplgulwtrtpkqnovngdbgiqjj6n2iefgkst6gdi
|
<start>The last two weeks of tweets from a safety researcher at OpenAI could be a science fiction short-short story.
(And, as usual, you can chose to believe or not whether OpenAI knows how to make an AI superintelligence, I have no special insight one way or another)
@emollick.bsky.social
|
The last two weeks of tweets from a safety researcher at OpenAI could be a science fiction short-short story.
(And, as usual, you can chose to believe or not whether OpenAI knows how to make an AI superintelligence, I have no special insight one way or another)
| 0 |
2025-01-15T17:02:57.835Z
|
2025-01-15T17:03:01.025Z
|
emollick.bsky.social
|
did:plc:flxq4uyjfotciovpw3x3fxnu
|
at://did:plc:flxq4uyjfotciovpw3x3fxnu/app.bsky.feed.post/3lfryqt26k226
|
bafyreiahtdbf32brbvjemvnkpp47rn4gmugxcdrjtbgwmob2k4e7hd5qla
|
<start>The national media found out about the hoverboards and they are concerned.
@emollick.bsky.social
|
The national media found out about the hoverboards and they are concerned.
| 0 |
2025-01-15T14:41:50.913Z
|
2025-01-15T14:41:51.917Z
|
emollick.bsky.social
|
did:plc:flxq4uyjfotciovpw3x3fxnu
|
at://did:plc:flxq4uyjfotciovpw3x3fxnu/app.bsky.feed.post/3lfqtbmekl226
|
bafyreifa6unboop6y4e254bw4e55w3jx2k3m4rtxwjxlc55zj72jyfb52q
|
<start>Lots of obsession over what to do as AI gets better, but I would love to see more attention to addressing some of the obvious and necessary problems we are encountering right now (which would make me feel better about our ability to handle even harder problems later).
@emollick.bsky.social
|
Lots of obsession over what to do as AI gets better, but I would love to see more attention to addressing some of the obvious and necessary problems we are encountering right now (which would make me feel better about our ability to handle even harder problems later).
| 0 |
2025-01-15T03:31:12.149Z
|
2025-01-15T03:31:12.818Z
|
emollick.bsky.social
|
did:plc:flxq4uyjfotciovpw3x3fxnu
|
at://did:plc:flxq4uyjfotciovpw3x3fxnu/app.bsky.feed.post/3lfqt6ruhrc26
|
bafyreibvpmxbsv5ag2fd2v4a67ho7nnnozhgbukr3bcelbzhtf5kf7xmlq
|
<start>Some near-term problems that need to be solved at current levels of AI capability:
1) AI audio and video and can be produced in real time. How do we establish identity & authenticity?
2) How do we manage the transformation of LLM-exposed jobs (writing, coding)?
3) What skills should we be teaching? Lots of obsession over what to do as AI gets better, but I would love to see more attention to addressing some of the obvious and necessary problems we are encountering right now (which would make me feel better about our ability to handle even harder problems later).
@emollick.bsky.social
|
Some near-term problems that need to be solved at current levels of AI capability:
1) AI audio and video and can be produced in real time. How do we establish identity & authenticity?
2) How do we manage the transformation of LLM-exposed jobs (writing, coding)?
3) What skills should we be teaching?
| 1 |
2025-01-15T03:29:37.250Z
|
2025-01-15T03:29:38.015Z
|
emollick.bsky.social
|
did:plc:flxq4uyjfotciovpw3x3fxnu
|
at://did:plc:flxq4uyjfotciovpw3x3fxnu/app.bsky.feed.post/3lfqgv6ves22t
|
bafyreiegd4vj6flmunuavlzwdgkudiq3o273jq6tcuuu7nbvz4wrjl3veu
|
<start>Bad new, they had to recall all the hoverboards, there was a design flaw. The national media found out about the hoverboards and they are concerned.
@emollick.bsky.social
|
Bad new, they had to recall all the hoverboards, there was a design flaw.
| 1 |
2025-01-14T23:49:30.465Z
|
2025-01-14T23:49:31.430Z
|
emollick.bsky.social
|
did:plc:flxq4uyjfotciovpw3x3fxnu
|
at://did:plc:flxq4uyjfotciovpw3x3fxnu/app.bsky.feed.post/3lfpskc6cgs2y
|
bafyreihhufkw4ivxb6sflhs4lpsbtrbnowqivwjbtnzz2ezsdmr2owcova
|
<start>Generative AI has flaws and biases, and there is a tendency for academics to fix on that (85% of equity LLM papers focus on harms)…
…yet in many ways LLMs are uniquely powerful among new technologies for helping people equitably in education and healthcare. We need an urgent focus on how to do that Paper: ai.nejm.org/doi/full/10....
@emollick.bsky.social
|
Generative AI has flaws and biases, and there is a tendency for academics to fix on that (85% of equity LLM papers focus on harms)…
…yet in many ways LLMs are uniquely powerful among new technologies for helping people equitably in education and healthcare. We need an urgent focus on how to do that
| 1 |
2025-01-14T17:45:29.965Z
|
2025-01-14T17:45:32.018Z
|
emollick.bsky.social
|
did:plc:flxq4uyjfotciovpw3x3fxnu
|
at://did:plc:flxq4uyjfotciovpw3x3fxnu/app.bsky.feed.post/3lfokw35xj22e
|
bafyreigbeg7y47dbs42nph33doaml6rkhegrcwmix52lsqjnhph3l5ti6m
|
<start>On one hand, this paper finds adding inference-time compute (like o1 does) improves medical reasoning, which is an important finding suggesting a way to continue to improve AI performance in medicine
On the other hand, scientific illustrations are apparently just anime now arxiv.org/pdf/2501.06458
@emollick.bsky.social
|
On one hand, this paper finds adding inference-time compute (like o1 does) improves medical reasoning, which is an important finding suggesting a way to continue to improve AI performance in medicine
On the other hand, scientific illustrations are apparently just anime now arxiv.org/pdf/2501.06458
| 0 |
2025-01-14T05:56:15.595Z
|
2025-01-14T05:56:18.120Z
|
emollick.bsky.social
|
did:plc:flxq4uyjfotciovpw3x3fxnu
|
at://did:plc:flxq4uyjfotciovpw3x3fxnu/app.bsky.feed.post/3lfohj7hfzs2e
|
bafyreibvjaju3tusil4hq4pkkrsnw767ce3y5j4hd6kp5flbqdjhqviiyy
|
<start>A few people said this wasn't realistic because there was no crowd. If you wanted a crowd you should have gone to the mech suit booth at CES (this video was produced on the first try of this prompt) Bad new, they had to recall all the hoverboards, there was a design flaw.
@emollick.bsky.social
|
A few people said this wasn't realistic because there was no crowd. If you wanted a crowd you should have gone to the mech suit booth at CES (this video was produced on the first try of this prompt)
| 1 |
2025-01-14T04:55:22.668Z
|
2025-01-14T04:55:24.222Z
|
emollick.bsky.social
|
did:plc:flxq4uyjfotciovpw3x3fxnu
|
at://did:plc:flxq4uyjfotciovpw3x3fxnu/app.bsky.feed.post/3lfo76z5zmc2x
|
bafyreig6tfxtmetkoz36syslvbrpbhdz6kqgqhbtfro5iut4mbu3kenxzi
|
<start>While I am doing hovering sports equipment, I might as well recreate another classic product. A few people said this wasn't realistic because there was no crowd. If you wanted a crowd you should have gone to the mech suit booth at CES (this video was produced on the first try of this prompt)
@emollick.bsky.social
|
While I am doing hovering sports equipment, I might as well recreate another classic product.
| 1 |
2025-01-14T02:26:30.590Z
|
2025-01-14T02:26:33.522Z
|
emollick.bsky.social
|
did:plc:flxq4uyjfotciovpw3x3fxnu
|
at://did:plc:flxq4uyjfotciovpw3x3fxnu/app.bsky.feed.post/3lfo6wmtquk2x
|
bafyreiayysz5kj3255gpcgctbqzbggatun4l6larntngndghjlwlqwowmu
|
<start>I don't think people know that making plausible fake videos is now something that doesn't require a lot of skill or time
I made this Veo video of a man trying out a flying skateboard at CES from a text prompt, 2nd try. Notice the shadows, green light, his body shifts for balance While I am doing hovering sports equipment, I might as well recreate another classic product.
@emollick.bsky.social
|
I don't think people know that making plausible fake videos is now something that doesn't require a lot of skill or time
I made this Veo video of a man trying out a flying skateboard at CES from a text prompt, 2nd try. Notice the shadows, green light, his body shifts for balance
| 1 |
2025-01-14T02:21:49.234Z
|
2025-01-14T02:21:52.227Z
|
emollick.bsky.social
|
did:plc:flxq4uyjfotciovpw3x3fxnu
|
at://did:plc:flxq4uyjfotciovpw3x3fxnu/app.bsky.feed.post/3lfn54u4fmk2q
|
bafyreiaqumjc7544glz34qdvtplu7fvs2s2toippfqtqx3qre2nq2zizvm
|
<start>Claude suggested Master and Command Gander when asked for some more puns (the rest were all mine), which was pretty good.
@emollick.bsky.social
|
Claude suggested Master and Command Gander when asked for some more puns (the rest were all mine), which was pretty good.
| 0 |
2025-01-13T16:16:50.962Z
|
2025-01-13T16:16:58.617Z
|
emollick.bsky.social
|
did:plc:flxq4uyjfotciovpw3x3fxnu
|
at://did:plc:flxq4uyjfotciovpw3x3fxnu/app.bsky.feed.post/3lfn2goni4c2q
|
bafyreigthv5lc6fr4y4mk77j2jsguysrrncpgxs2ges46nsl2z5lsejx2a
|
<start>“Instructors should persuade students that they are benefitting... [or they] can be misled by the inherent disfluency associated with the sustained cognitive effort required for active learning, which...can have a negative impact on their actual learning.” www.pnas.org/content/pnas...
@emollick.bsky.social
|
“Instructors should persuade students that they are benefitting... [or they] can be misled by the inherent disfluency associated with the sustained cognitive effort required for active learning, which...can have a negative impact on their actual learning.” www.pnas.org/content/pnas...
| 0 |
2025-01-13T15:28:39.546Z
|
2025-01-13T15:28:46.627Z
|
emollick.bsky.social
|
did:plc:flxq4uyjfotciovpw3x3fxnu
|
at://did:plc:flxq4uyjfotciovpw3x3fxnu/app.bsky.feed.post/3lfn2dg5zus2q
|
bafyreibvjg45cfiddsowa5gobtjojm65mgrtvttwjqrm5d5wcgw4rnse4e
|
<start>I think often about this experiment at Harvard showing active learning upped test scores by 33%... but students thought they were learning more from non-active lectures
The problem: being challenged is the key to learning BUT we hate challenges, they are hard & make you realize how little you know. “Instructors should persuade students that they are benefitting... [or they] can be misled by the inherent disfluency associated with the sustained cognitive effort required for active learning, which...can have a negative impact on their actual learning.” www.pnas.org/content/pnas...
@emollick.bsky.social
|
I think often about this experiment at Harvard showing active learning upped test scores by 33%... but students thought they were learning more from non-active lectures
The problem: being challenged is the key to learning BUT we hate challenges, they are hard & make you realize how little you know.
| 1 |
2025-01-13T15:26:49.986Z
|
2025-01-13T15:27:04.625Z
|
emollick.bsky.social
|
did:plc:flxq4uyjfotciovpw3x3fxnu
|
at://did:plc:flxq4uyjfotciovpw3x3fxnu/app.bsky.feed.post/3lfmwiet4xk2q
|
bafyreicfnc23zqbqkvfkf6e3vlpjvdrwnl2r5svj3gizpxfhhouw65jtga
|
<start>I also agree with @simonwillison.net that the labs should be clearly publishing this data, since environmental impact remains one of the most common, confusing and misleading debates about the impact of AI.
@emollick.bsky.social
|
I also agree with @simonwillison.net that the labs should be clearly publishing this data, since environmental impact remains one of the most common, confusing and misleading debates about the impact of AI.
| 0 |
2025-01-13T14:18:01.390Z
|
2025-01-13T14:18:01.917Z
|
emollick.bsky.social
|
did:plc:flxq4uyjfotciovpw3x3fxnu
|
at://did:plc:flxq4uyjfotciovpw3x3fxnu/app.bsky.feed.post/3lfm4zdi6222x
|
bafyreia4nz2g3ye4uriq2rtlzznfjnnc6ugdmkjl3u2wycaliakwjqneyi
|
<start>Some evidence of further energy improvements (10x!) on Llama 3.3 70B. bsky.app/profile/lhl.... I also agree with @simonwillison.net that the labs should be clearly publishing this data, since environmental impact remains one of the most common, confusing and misleading debates about the impact of AI.
@emollick.bsky.social
|
Some evidence of further energy improvements (10x!) on Llama 3.3 70B. bsky.app/profile/lhl....
| 1 |
2025-01-13T06:42:13.121Z
|
2025-01-13T06:42:13.417Z
|
emollick.bsky.social
|
did:plc:flxq4uyjfotciovpw3x3fxnu
|
at://did:plc:flxq4uyjfotciovpw3x3fxnu/app.bsky.feed.post/3lfltmoba2s2x
|
bafyreigch2hse22ag6i46fdle2azwgcqnmcj2y7astnamjbkdsutapjg3a
|
<start>Looks like he is using the 2020 GPT-3 numbers I referred to in the post, not measuring energy use directly.
@emollick.bsky.social
|
Looks like he is using the 2020 GPT-3 numbers I referred to in the post, not measuring energy use directly.
| 0 |
2025-01-13T03:54:04.544Z
|
2025-01-13T03:54:05.120Z
|
emollick.bsky.social
|
did:plc:flxq4uyjfotciovpw3x3fxnu
|
at://did:plc:flxq4uyjfotciovpw3x3fxnu/app.bsky.feed.post/3lflsqt7nik2x
|
bafyreialqzgctrt5nis6lejne2xen2o3nri73ioiahhbkpdb37udpre4gy
|
<start>If you are going to reply with AI slop to a post, at least make it interesting. I am getting sick of the "fascinating insight!" comments flooding my replies.
You can do so much better, here are some suggestions.
@emollick.bsky.social
|
If you are going to reply with AI slop to a post, at least make it interesting. I am getting sick of the "fascinating insight!" comments flooding my replies.
You can do so much better, here are some suggestions.
| 0 |
2025-01-13T03:38:30.201Z
|
2025-01-13T03:38:32.925Z
|
emollick.bsky.social
|
did:plc:flxq4uyjfotciovpw3x3fxnu
|
at://did:plc:flxq4uyjfotciovpw3x3fxnu/app.bsky.feed.post/3lflrpxqmg22x
|
bafyreicbervnq2dxmcuilmzsklabckzjtj4pdpmluxkqznfxpzawwjwf7u
|
<start>Conference papers typically go through review (though I think your opinion of what peer review catches is a little too optimistic), but the good thing about this particular paper is that all of the data is public using an open weights model. It also matches other calculations based on GPU usage.
@emollick.bsky.social
|
Conference papers typically go through review (though I think your opinion of what peer review catches is a little too optimistic), but the good thing about this particular paper is that all of the data is public using an open weights model. It also matches other calculations based on GPU usage.
| 0 |
2025-01-13T03:20:07.664Z
|
2025-01-13T03:20:08.116Z
|
emollick.bsky.social
|
did:plc:flxq4uyjfotciovpw3x3fxnu
|
at://did:plc:flxq4uyjfotciovpw3x3fxnu/app.bsky.feed.post/3lflrefjwo22x
|
bafyreic67twpfi7nxvkxlmoiacqibjph6wv3wlp5xk5vd4rl5zqvw3cehi
|
<start>(a token is roughly a word in length). Some evidence of further energy improvements (10x!) on Llama 3.3 70B. bsky.app/profile/lhl....
@emollick.bsky.social
|
(a token is roughly a word in length).
| 1 |
2025-01-13T03:13:39.473Z
|
2025-01-13T03:13:39.720Z
|
emollick.bsky.social
|
did:plc:flxq4uyjfotciovpw3x3fxnu
|
at://did:plc:flxq4uyjfotciovpw3x3fxnu/app.bsky.feed.post/3lflqvft53c2x
|
bafyreiamgqrft7stdhdg3g2ktg64rlp6poftklaaaidvpwrqzv6pgy2mpi
|
<start>o1 will almost certainly use a lot more energy per (output) token, GPT-4o-mini or Gemini Flash or the other free models will use a lot less.
And, of course, in aggregate AI data centers will have a real environmental impact. (a token is roughly a word in length).
@emollick.bsky.social
|
o1 will almost certainly use a lot more energy per (output) token, GPT-4o-mini or Gemini Flash or the other free models will use a lot less.
And, of course, in aggregate AI data centers will have a real environmental impact.
| 1 |
2025-01-13T03:05:16.458Z
|
2025-01-13T03:05:17.020Z
|
emollick.bsky.social
|
did:plc:flxq4uyjfotciovpw3x3fxnu
|
at://did:plc:flxq4uyjfotciovpw3x3fxnu/app.bsky.feed.post/3lflq4rwnbk2x
|
bafyreihpnucg5nlapn5rqtvplt4m7dwtg6zkh7q6kyp6gtt65digtump2e
|
<start>This does not count training, which was estimated at a little above 500,000 kWh, about 18 hours of a Boeing 737 in flight (or 571 years of a TV streaming Netflix) o1 will almost certainly use a lot more energy per (output) token, GPT-4o-mini or Gemini Flash or the other free models will use a lot less.
And, of course, in aggregate AI data centers will have a real environmental impact.
@emollick.bsky.social
|
This does not count training, which was estimated at a little above 500,000 kWh, about 18 hours of a Boeing 737 in flight (or 571 years of a TV streaming Netflix)
| 1 |
2025-01-13T02:51:30.295Z
|
2025-01-13T02:51:30.815Z
|
emollick.bsky.social
|
did:plc:flxq4uyjfotciovpw3x3fxnu
|
at://did:plc:flxq4uyjfotciovpw3x3fxnu/app.bsky.feed.post/3lflpodl3bc2x
|
bafyreicbrcs24dllfax4krmjlwkdapo3hrx66fxc2xpjyxnaksyj3brlf4
|
<start>Most of the talk around AI and energy use refer to an older 2020 estimate of GPT-3 energy consumption, but a more recent paper directly measures energy use of Llama 65B as 3-4 joules per decoded token.
So an hour of streaming Netflix is equivalent to 70-90,000 65B tokens. arxiv.org/pdf/2310.03003 This does not count training, which was estimated at a little above 500,000 kWh, about 18 hours of a Boeing 737 in flight (or 571 years of a TV streaming Netflix)
@emollick.bsky.social
|
Most of the talk around AI and energy use refer to an older 2020 estimate of GPT-3 energy consumption, but a more recent paper directly measures energy use of Llama 65B as 3-4 joules per decoded token.
So an hour of streaming Netflix is equivalent to 70-90,000 65B tokens. arxiv.org/pdf/2310.03003
| 1 |
2025-01-13T02:43:25.464Z
|
2025-01-13T02:43:27.621Z
|
emollick.bsky.social
|
did:plc:flxq4uyjfotciovpw3x3fxnu
|
at://did:plc:flxq4uyjfotciovpw3x3fxnu/app.bsky.feed.post/3lflmzox7dk2x
|
bafyreihy4gdvmm6d7ionuuxu4dfgklvantc2fmwlifu4zgcd77gpjlxvnu
|
<start>We are actively experimenting with using them in our educational simulations. You can read more about those in this paper: papers.ssrn.com/sol3/papers....
@emollick.bsky.social
|
We are actively experimenting with using them in our educational simulations. You can read more about those in this paper: papers.ssrn.com/sol3/papers....
| 0 |
2025-01-13T01:56:05.279Z
|
2025-01-13T01:56:05.621Z
|
emollick.bsky.social
|
did:plc:flxq4uyjfotciovpw3x3fxnu
|
at://did:plc:flxq4uyjfotciovpw3x3fxnu/app.bsky.feed.post/3lflmxljt7c2x
|
bafyreiedlbmxguanbm443i2bqltbzmi5qov7ifgcjzzfjm7syi4quyqnai
|
<start>You could take a look at our publications on AI in education and work and see that we are doing exactly that, including open prompt libraries for teaching.
And yes, aggregate use of energy by AI is a concern, but querying ChatGPT uses less energy than streaming a TV show for the same time period.
@emollick.bsky.social
|
You could take a look at our publications on AI in education and work and see that we are doing exactly that, including open prompt libraries for teaching.
And yes, aggregate use of energy by AI is a concern, but querying ChatGPT uses less energy than streaming a TV show for the same time period.
| 0 |
2025-01-13T01:54:54.586Z
|
2025-01-13T01:54:55.127Z
|
Subsets and Splits
No community queries yet
The top public SQL queries from the community will appear here once available.