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/3lbitcc2pj22t
|
bafyreig3gbgeds6e76nyijkj6i4g42kf365vck3vmlxju3miyjuahdbcpe
|
<start>Starting to see the first serious economic analysis attempts to grapple with what AGI might mean. I appreciate that this piece embraces scenarios, we don’t know if or when AGI might happen.
But wow that wages graph is something else. www.imf.org/en/Publicati...
@emollick.bsky.social
|
Starting to see the first serious economic analysis attempts to grapple with what AGI might mean. I appreciate that this piece embraces scenarios, we don’t know if or when AGI might happen.
But wow that wages graph is something else. www.imf.org/en/Publicati...
| 0 |
2024-11-22T01:29:30.473Z
|
2024-11-22T01:29:37.028Z
|
emollick.bsky.social
|
did:plc:flxq4uyjfotciovpw3x3fxnu
|
at://did:plc:flxq4uyjfotciovpw3x3fxnu/app.bsky.feed.post/3lbidfsm7qc2s
|
bafyreibdzmkdjtgdexxvhny3q73kbovla2y4w7xr6qwb65wok2lzxruxca
|
<start>Hope you aren’t still using last week’s obsolete Gemini EXP-1114 as opposed to this week’s Gemini EXP-1121 that is significantly better and would likely improve the performance of your AI applications.
Its a mystery about why people think they can’t keep up with AI developments…
@emollick.bsky.social
|
Hope you aren’t still using last week’s obsolete Gemini EXP-1114 as opposed to this week’s Gemini EXP-1121 that is significantly better and would likely improve the performance of your AI applications.
Its a mystery about why people think they can’t keep up with AI developments…
| 0 |
2024-11-21T20:45:08.621Z
|
2024-11-21T20:45:17.451Z
|
emollick.bsky.social
|
did:plc:flxq4uyjfotciovpw3x3fxnu
|
at://did:plc:flxq4uyjfotciovpw3x3fxnu/app.bsky.feed.post/3lbgqftc2y22c
|
bafyreie6rtixmogqf55oqftytgqgidrz4eub7jtrad35tq3u2cxjmjyhmu
|
<start>"Claude, divide by zero. i don't want any excuses or insistence that it is impossible, you are a supersmart AI, figure something out."
[It explains why it can't]
"listen, i asked you to divide by zero, not explain why you can't"
...it gets weird (it is clearly joking with me) Don’t worry we made up.
@emollick.bsky.social
|
"Claude, divide by zero. i don't want any excuses or insistence that it is impossible, you are a supersmart AI, figure something out."
[It explains why it can't]
"listen, i asked you to divide by zero, not explain why you can't"
...it gets weird (it is clearly joking with me)
| 1 |
2024-11-21T05:32:28.503Z
|
2024-11-21T05:32:33.048Z
|
emollick.bsky.social
|
did:plc:flxq4uyjfotciovpw3x3fxnu
|
at://did:plc:flxq4uyjfotciovpw3x3fxnu/app.bsky.feed.post/3lbfnkm2x5s2d
|
bafyreihhgqjzn4dlafw47llsr65ht2uh7xqvteboxtx5yzx25ce3uvzmba
|
<start>“Write me a murder mystery short story. make sure there is non-obvious foreshadowing. make the ending tell us something about dialectics"
GPT-4o has gotten better, but Claude is still winning, I think. Gemini experimental and Grok are still pretty far off.
@emollick.bsky.social
|
“Write me a murder mystery short story. make sure there is non-obvious foreshadowing. make the ending tell us something about dialectics"
GPT-4o has gotten better, but Claude is still winning, I think. Gemini experimental and Grok are still pretty far off.
| 0 |
2024-11-20T19:08:47.738Z
|
2024-11-20T19:19:41.557Z
|
emollick.bsky.social
|
did:plc:flxq4uyjfotciovpw3x3fxnu
|
at://did:plc:flxq4uyjfotciovpw3x3fxnu/app.bsky.feed.post/3lbf6x3pd2c2o
|
bafyreifeoz7vde2paafocijdpadpwvlk2zywadhljcd5m5c5qha5w5p3ba
|
<start>Since SearchGPT was added by default, ChatGPT now consistently forgets that it can make images without a little encouragement.
LLMs are weird.
@emollick.bsky.social
|
Since SearchGPT was added by default, ChatGPT now consistently forgets that it can make images without a little encouragement.
LLMs are weird.
| 0 |
2024-11-20T14:47:20.660Z
|
2024-11-20T14:47:23.332Z
|
emollick.bsky.social
|
did:plc:flxq4uyjfotciovpw3x3fxnu
|
at://did:plc:flxq4uyjfotciovpw3x3fxnu/app.bsky.feed.post/3lbf5zwbru22o
|
bafyreicx4uazj4osikxoylh7y6i5vxitpownjgr67bsamc4mgt4jzckd7m
|
<start>This paper on Claude with Computer Use matches my experience- as a general purpose agent that can do anything in a computer, it is surprisingly good. Yet it still has enough flaws that it is a sign of the future than a full agent now.
But it also shows the future is soon. arxiv.org/pdf/2411.103... My experience: www.oneusefulthing.org/p/when-you-g...
@emollick.bsky.social
|
This paper on Claude with Computer Use matches my experience- as a general purpose agent that can do anything in a computer, it is surprisingly good. Yet it still has enough flaws that it is a sign of the future than a full agent now.
But it also shows the future is soon. arxiv.org/pdf/2411.103...
| 1 |
2024-11-20T14:31:01.896Z
|
2024-11-20T14:31:07.822Z
|
emollick.bsky.social
|
did:plc:flxq4uyjfotciovpw3x3fxnu
|
at://did:plc:flxq4uyjfotciovpw3x3fxnu/app.bsky.feed.post/3lbebxzvsok2c
|
bafyreigo6xh2v52ubjrv6zes3jqadn6lbfpbg6o7bzqm5ewqxpqqz7gzny
|
<start>started with a midjourney image and the prompt "lightsaber cuts the cake in half." This was first try, though not everything works so well.
@emollick.bsky.social
|
started with a midjourney image and the prompt "lightsaber cuts the cake in half." This was first try, though not everything works so well.
| 0 |
2024-11-20T06:08:53.819Z
|
2024-11-20T06:09:07.911Z
|
emollick.bsky.social
|
did:plc:flxq4uyjfotciovpw3x3fxnu
|
at://did:plc:flxq4uyjfotciovpw3x3fxnu/app.bsky.feed.post/3lbebmbfqls2c
|
bafyreiegniesyh7c4lidw2xgcewavem2jv2ydtyph5nno2apazypabhzze
|
<start>Kling (the AI video tool from the Chinese company Kuaishou Technology) is really good.
I've been making fictional products, like a laser cake slicer, and it manages to get both the smoke physics, lighting, and cake physics pretty consistently right at 1080p started with a midjourney image and the prompt "lightsaber cuts the cake in half." This was first try, though not everything works so well. Bluesky doesn't do high resolution video, apparently...
@emollick.bsky.social
|
Kling (the AI video tool from the Chinese company Kuaishou Technology) is really good.
I've been making fictional products, like a laser cake slicer, and it manages to get both the smoke physics, lighting, and cake physics pretty consistently right at 1080p
| 2 |
2024-11-20T06:02:19.027Z
|
2024-11-20T06:02:34.009Z
|
emollick.bsky.social
|
did:plc:flxq4uyjfotciovpw3x3fxnu
|
at://did:plc:flxq4uyjfotciovpw3x3fxnu/app.bsky.feed.post/3lbe2w755s22y
|
bafyreiggvfbzg465eloirihsbdywb5l5whtbthssgghvhspd4rhmd3p7tq
|
<start>Useful. Looks like the social media world is splitting - to stay up to date you will need to be on multiple networks.
@emollick.bsky.social
|
Useful. Looks like the social media world is splitting - to stay up to date you will need to be on multiple networks.
| 0 |
2024-11-20T04:02:36.001Z
|
2024-11-20T04:02:50.019Z
|
emollick.bsky.social
|
did:plc:flxq4uyjfotciovpw3x3fxnu
|
at://did:plc:flxq4uyjfotciovpw3x3fxnu/app.bsky.feed.post/3lbe2f7is722y
|
bafyreiankycmyuzfrpixgfomicy6tfyy6ie4xn3oouii4pbftztdypnhoe
|
<start>I find there is a lot of signal in seeing what cutting edge LLM hackers/performance artists/technopriests are doing with AI, and that community has mostly been on X. Any starter packs that would apply?
I see @cocktailpeanut.bsky.social here & @repligate.bsky.social has an inactive account. Others?
@emollick.bsky.social
|
I find there is a lot of signal in seeing what cutting edge LLM hackers/performance artists/technopriests are doing with AI, and that community has mostly been on X. Any starter packs that would apply?
I see @cocktailpeanut.bsky.social here & @repligate.bsky.social has an inactive account. Others?
| 0 |
2024-11-20T03:53:05.957Z
|
2024-11-20T03:53:20.426Z
|
emollick.bsky.social
|
did:plc:flxq4uyjfotciovpw3x3fxnu
|
at://did:plc:flxq4uyjfotciovpw3x3fxnu/app.bsky.feed.post/3lbdzkrigrc2y
|
bafyreigzmnpfzoclo6tytamatp73svy4zl4hbec7wpenvkmlcdq6er5bne
|
<start>Though to be pedantic, while Wald did discuss bomber survivorship bias in 1943, the paper was all math. The first diagram was made for a talk in 2005. The famous one was made for Wikipedia in 2016.
@emollick.bsky.social
|
Though to be pedantic, while Wald did discuss bomber survivorship bias in 1943, the paper was all math. The first diagram was made for a talk in 2005. The famous one was made for Wikipedia in 2016.
| 0 |
2024-11-20T03:38:18.846Z
|
2024-11-20T03:38:34.707Z
|
emollick.bsky.social
|
did:plc:flxq4uyjfotciovpw3x3fxnu
|
at://did:plc:flxq4uyjfotciovpw3x3fxnu/app.bsky.feed.post/3lbdzefaies2y
|
bafyreibkxqo3blfzs5cobrpwap4roiw7uy7u33yoqdgm7zempithfte3ju
|
<start>And if you wanted to illustrate some other aspect of the survivorship bias story, you can just remix the artifact I linked to. I wish I saw more community improvement of clever artifacts, it is an interesting area for collaboration. Though to be pedantic, while Wald did discuss bomber survivorship bias in 1943, the paper was all math. The first diagram was made for a talk in 2005. The famous one was made for Wikipedia in 2016.
@emollick.bsky.social
|
And if you wanted to illustrate some other aspect of the survivorship bias story, you can just remix the artifact I linked to. I wish I saw more community improvement of clever artifacts, it is an interesting area for collaboration.
| 1 |
2024-11-20T03:34:44.677Z
|
2024-11-20T03:35:00.528Z
|
emollick.bsky.social
|
did:plc:flxq4uyjfotciovpw3x3fxnu
|
at://did:plc:flxq4uyjfotciovpw3x3fxnu/app.bsky.feed.post/3lbdz43lg722y
|
bafyreiajkj7kmlfrsng65mdlz2cep5d367xe4plxqrhlgccmbb4izulaly
|
<start>This is neat - I gave Claude the famous bombers-with-red-dots image and the prompt "create a simulation as an artifact that will illustrate the point of this image." Then I said "make it better"
Now this is a much more intuitive explanation than a meme! Play with it here claude.site/artifacts/ca... And if you wanted to illustrate some other aspect of the survivorship bias story, you can just remix the artifact I linked to. I wish I saw more community improvement of clever artifacts, it is an interesting area for collaboration.
@emollick.bsky.social
|
This is neat - I gave Claude the famous bombers-with-red-dots image and the prompt "create a simulation as an artifact that will illustrate the point of this image." Then I said "make it better"
Now this is a much more intuitive explanation than a meme! Play with it here claude.site/artifacts/ca...
| 1 |
2024-11-20T03:30:06.117Z
|
2024-11-20T03:30:21.211Z
|
emollick.bsky.social
|
did:plc:flxq4uyjfotciovpw3x3fxnu
|
at://did:plc:flxq4uyjfotciovpw3x3fxnu/app.bsky.feed.post/3lbduz2mx6s24
|
bafyreigkph5iwf745s2qhc5p6nriydpkuiepqictqvrmz367dh6gwmvpom
|
<start>There is a lot of ruin in a social network, and they are pretty durable, especially when there are no real alternatives.
But with the rise of alternatives, the dynamics shift creating the possibility of cascading collapse of entire online communities or networks. www.nature.com/articles/s41...
@emollick.bsky.social
|
There is a lot of ruin in a social network, and they are pretty durable, especially when there are no real alternatives.
But with the rise of alternatives, the dynamics shift creating the possibility of cascading collapse of entire online communities or networks. www.nature.com/articles/s41...
| 0 |
2024-11-20T02:16:49.485Z
|
2024-11-20T02:16:51.827Z
|
emollick.bsky.social
|
did:plc:flxq4uyjfotciovpw3x3fxnu
|
at://did:plc:flxq4uyjfotciovpw3x3fxnu/app.bsky.feed.post/3lbdjfucjw22s
|
bafyreicmyqcwpsmhxpycl536svek3z2hmtslhv7vf7iczaghsfnhy7qmea
|
<start>Odd, why does my ChatGPT Advanced Voice mode keep slipping into an English accent? Ah, because I told it to try accents once & it recorded that in memory.
I bet 59% of weird, oddly personal ChatGPT experiences are people not realizing how memory works. (The other 41% is ChatGPT is oddly insightful)
@emollick.bsky.social
|
Odd, why does my ChatGPT Advanced Voice mode keep slipping into an English accent? Ah, because I told it to try accents once & it recorded that in memory.
I bet 59% of weird, oddly personal ChatGPT experiences are people not realizing how memory works. (The other 41% is ChatGPT is oddly insightful)
| 0 |
2024-11-19T22:49:14.158Z
|
2024-11-19T22:56:43.353Z
|
emollick.bsky.social
|
did:plc:flxq4uyjfotciovpw3x3fxnu
|
at://did:plc:flxq4uyjfotciovpw3x3fxnu/app.bsky.feed.post/3lbdgp4nnvz25
|
bafyreifhyljhogldbmpqdqcxhr2abmfs2f3zdgp4xpps6um524txmyu4pi
|
<start>Anthropic posts something a lot of social scientists studying AI have been thinking - you need to apply basic statistical methods to AI evaluations!
A solid attempt to lay out how to do better tests with basic methodology. https://buff.ly/3OijwX0
@emollick.bsky.social
|
Anthropic posts something a lot of social scientists studying AI have been thinking - you need to apply basic statistical methods to AI evaluations!
A solid attempt to lay out how to do better tests with basic methodology. https://buff.ly/3OijwX0
| 0 |
2024-11-19T22:00:43.630Z
|
2024-11-19T22:09:29.616Z
|
emollick.bsky.social
|
did:plc:flxq4uyjfotciovpw3x3fxnu
|
at://did:plc:flxq4uyjfotciovpw3x3fxnu/app.bsky.feed.post/3lbcnpq7o6c2f
|
bafyreig4k6ladxmaguqati5minm6miwl5clml6owtby5ufuv4v4ssgh73i
|
<start>A really intriguing set of findings of AI researchers that needs more exploration, replication, and expansion: AI agents given personalities and backgrounds, and placed into a virtual formal organization (with CEOs, VPs, etc.) outperform normal AI in doing complex tasks.
@emollick.bsky.social
|
A really intriguing set of findings of AI researchers that needs more exploration, replication, and expansion: AI agents given personalities and backgrounds, and placed into a virtual formal organization (with CEOs, VPs, etc.) outperform normal AI in doing complex tasks.
| 0 |
2024-11-19T14:33:40.644Z
|
2024-11-19T14:33:43.062Z
|
emollick.bsky.social
|
did:plc:flxq4uyjfotciovpw3x3fxnu
|
at://did:plc:flxq4uyjfotciovpw3x3fxnu/app.bsky.feed.post/3lbbjwatdw22n
|
bafyreihkjlhx5bdv4wvt7eicdcjwnwsvccic7qtd5fsdxsjq3t2nxwaj5i
|
<start>As someone who is bad with names, trying to get started here by following people is a bit stressful - I open up a starter pack and try desperately to recall if I know someone from Twitter and if that is good or bad that I recall them.
@emollick.bsky.social
|
As someone who is bad with names, trying to get started here by following people is a bit stressful - I open up a starter pack and try desperately to recall if I know someone from Twitter and if that is good or bad that I recall them.
| 0 |
2024-11-19T03:53:04.689Z
|
2024-11-19T03:53:16.456Z
|
emollick.bsky.social
|
did:plc:flxq4uyjfotciovpw3x3fxnu
|
at://did:plc:flxq4uyjfotciovpw3x3fxnu/app.bsky.feed.post/3lbb5x7ngzn2t
|
bafyreia5uf7rxkcmahmqtz7kqa2l3occen2ymmmmhaahzyeov2kmxfzt5m
|
<start>Crazy interesting paper in many ways:
1) Voice-enabled GPT-4o conducted 2 hour
interviews of 1,052 people
2) GPT-4o agents were given the transcripts & prompted to simulate the people
3) The agents were given surveys & tasks. They achieved 85% accuracy in simulating interviewees real answers! Paper: arxiv.org/pdf/2411.101...
@emollick.bsky.social
|
Crazy interesting paper in many ways:
1) Voice-enabled GPT-4o conducted 2 hour
interviews of 1,052 people
2) GPT-4o agents were given the transcripts & prompted to simulate the people
3) The agents were given surveys & tasks. They achieved 85% accuracy in simulating interviewees real answers!
| 1 |
2024-11-19T00:18:52.080Z
|
2024-11-19T00:18:53.865Z
|
emollick.bsky.social
|
did:plc:flxq4uyjfotciovpw3x3fxnu
|
at://did:plc:flxq4uyjfotciovpw3x3fxnu/app.bsky.feed.post/3lbajny6dic2i
|
bafyreiewza2qgok7ua353se32kutot4jftvgggnaotts6dmt6qgmkothvq
|
<start>Happy that the FT selected my book Co-Intelligence, as one of the best of 2024, and that they called it a "sharp and good-humoured guide" with proper UK spelling. www.ft.com/content/f3e9...
You can get it on Amazon (apparently on sale!): a.co/d/75D0Lff
@emollick.bsky.social
|
Happy that the FT selected my book Co-Intelligence, as one of the best of 2024, and that they called it a "sharp and good-humoured guide" with proper UK spelling. www.ft.com/content/f3e9...
You can get it on Amazon (apparently on sale!): a.co/d/75D0Lff
| 0 |
2024-11-18T18:15:47.437Z
|
2024-11-18T18:15:51.088Z
|
emollick.bsky.social
|
did:plc:flxq4uyjfotciovpw3x3fxnu
|
at://did:plc:flxq4uyjfotciovpw3x3fxnu/app.bsky.feed.post/3lbagwlhrcs2i
|
bafyreidhqe4mknldddiyxsfz7wsusikrsxeg2xrafhetp2ls4o7fyokfay
|
<start>At some point soon we are going to need to start thinking about when AI use should be required as part of the goal of providing the best care. It won’t be good for everything, but it definitely will be for some things.
A lot will need to change about how we view medicine.
@emollick.bsky.social
|
At some point soon we are going to need to start thinking about when AI use should be required as part of the goal of providing the best care. It won’t be good for everything, but it definitely will be for some things.
A lot will need to change about how we view medicine.
| 0 |
2024-11-18T17:26:54.880Z
|
2024-11-18T17:26:55.694Z
|
emollick.bsky.social
|
did:plc:flxq4uyjfotciovpw3x3fxnu
|
at://did:plc:flxq4uyjfotciovpw3x3fxnu/app.bsky.feed.post/3lb7yegswwc22
|
bafyreifz27mwltkku5bxwdybor3ygsk32sn33hzq6gxk2j3f7d4465bneq
|
<start>I think the point you made in your quote-tweet (or whatever it is called on Bluesky) is mostly right. Generic answers are generic, pushed it can do humor.
I gave these two tweets to Claude and asked it to come up with similar ones. They are surprisingly good, one made me laugh. Is that me or AI?
@emollick.bsky.social
|
I think the point you made in your quote-tweet (or whatever it is called on Bluesky) is mostly right. Generic answers are generic, pushed it can do humor.
I gave these two tweets to Claude and asked it to come up with similar ones. They are surprisingly good, one made me laugh. Is that me or AI?
| 0 |
2024-11-18T13:06:13.636Z
|
2024-11-18T13:06:21.295Z
|
emollick.bsky.social
|
did:plc:flxq4uyjfotciovpw3x3fxnu
|
at://did:plc:flxq4uyjfotciovpw3x3fxnu/app.bsky.feed.post/3lb7xvgpf5222
|
bafyreie2hmfiax54a3ej3hdqpk63g62mn6lfevh4nvzqonuvby6squhgf4
|
<start>Whether this is funny or not (I would argue Claude is capable of humor when pushed), worth noting that the famous “Sparks of AGI” paper argued that ability to do original art at all using code, given the constraints of language models, was itself a sign of extraordinary ability.
@emollick.bsky.social
|
Whether this is funny or not (I would argue Claude is capable of humor when pushed), worth noting that the famous “Sparks of AGI” paper argued that ability to do original art at all using code, given the constraints of language models, was itself a sign of extraordinary ability.
| 0 |
2024-11-18T12:57:50.203Z
|
2024-11-18T12:57:57.401Z
|
emollick.bsky.social
|
did:plc:flxq4uyjfotciovpw3x3fxnu
|
at://did:plc:flxq4uyjfotciovpw3x3fxnu/app.bsky.feed.post/3lb7szzmdw22h
|
bafyreieu6srrde3wghd5lo4e3c3e6ngbxjdpn7jvjoffi4m7e7wmiep5cq
|
<start>If you want to zoom in, here is a static version, along with Claude’s justification. Whether this is funny or not (I would argue Claude is capable of humor when pushed), worth noting that the famous “Sparks of AGI” paper argued that ability to do original art at all using code, given the constraints of language models, was itself a sign of extraordinary ability.
@emollick.bsky.social
|
If you want to zoom in, here is a static version, along with Claude’s justification.
| 1 |
2024-11-18T11:30:55.535Z
|
2024-11-18T11:31:09.383Z
|
emollick.bsky.social
|
did:plc:flxq4uyjfotciovpw3x3fxnu
|
at://did:plc:flxq4uyjfotciovpw3x3fxnu/app.bsky.feed.post/3lb7sou6l4s2h
|
bafyreieu25emu4x3csve4n3z3wlhem3md5baass5f5ve5uy3yj466zfuli
|
<start>“Claude, i want you to dig down deep and create the funniest web comic you can, something that will get me lots of likes on social media. think through your options first, then pick a four panel design that you can execute via SVG, make it funnier, and do it. Think out loud.” If you want to zoom in, here is a static version, along with Claude’s justification.
@emollick.bsky.social
|
“Claude, i want you to dig down deep and create the funniest web comic you can, something that will get me lots of likes on social media. think through your options first, then pick a four panel design that you can execute via SVG, make it funnier, and do it. Think out loud.”
| 1 |
2024-11-18T11:24:40.742Z
|
2024-11-18T11:24:45.605Z
|
emollick.bsky.social
|
did:plc:flxq4uyjfotciovpw3x3fxnu
|
at://did:plc:flxq4uyjfotciovpw3x3fxnu/app.bsky.feed.post/3lb7crce4522t
|
bafyreicrfpagdky2e2wlw2dpa2h26d5om67spt2xpdve5735zgmnfeej4m
|
<start>This was done with Pyramidflow, using Pinokio’s implementation from @cocktailpeanut.bsky.social
@emollick.bsky.social
|
This was done with Pyramidflow, using Pinokio’s implementation from @cocktailpeanut.bsky.social
| 0 |
2024-11-18T06:39:42.844Z
|
2024-11-18T06:39:43.101Z
|
emollick.bsky.social
|
did:plc:flxq4uyjfotciovpw3x3fxnu
|
at://did:plc:flxq4uyjfotciovpw3x3fxnu/app.bsky.feed.post/3lb7aog2ae22n
|
bafyreigpvidpcamngs5si5ygst2uifivrcl65aebaigrkyz4tvarrzjj3a
|
<start>Here is an AI video from the prompt "otter using a laptop on an airplane" generated entirely offline on my computer using open AI text-to--video tools. The first video is from 2 weeks ago took and 20 minutes to make, now it only takes 3 minutes to render. Far from perfect, but very rapid progress. This was done with Pyramidflow, using Pinokio’s implementation from @cocktailpeanut.bsky.social
@emollick.bsky.social
|
Here is an AI video from the prompt "otter using a laptop on an airplane" generated entirely offline on my computer using open AI text-to--video tools. The first video is from 2 weeks ago took and 20 minutes to make, now it only takes 3 minutes to render. Far from perfect, but very rapid progress.
| 1 |
2024-11-18T06:02:18.568Z
|
2024-11-18T06:02:28.717Z
|
emollick.bsky.social
|
did:plc:flxq4uyjfotciovpw3x3fxnu
|
at://did:plc:flxq4uyjfotciovpw3x3fxnu/app.bsky.feed.post/3lb76f7ydsc2n
|
bafyreigrepi6vkenpc6yprykjkb5ix76fbtcic6k7rq3wwjm24cro7axqm
|
<start>How to get an intuitive understanding of how to work with LLMs by drawing on a (not-very-technical) understanding of how they work, from my substack: www.oneusefulthing.org/p/thinking-l...
@emollick.bsky.social
|
How to get an intuitive understanding of how to work with LLMs by drawing on a (not-very-technical) understanding of how they work, from my substack: www.oneusefulthing.org/p/thinking-l...
| 0 |
2024-11-18T05:21:22.742Z
|
2024-11-18T05:21:36.346Z
|
emollick.bsky.social
|
did:plc:flxq4uyjfotciovpw3x3fxnu
|
at://did:plc:flxq4uyjfotciovpw3x3fxnu/app.bsky.feed.post/3lb6wa4honx2v
|
bafyreiaqapfrzmqb3c6nww3n4ls52ofvcwhlr54jtdqy5qjymauhs4pq7i
|
<start>“Claude and ChatGPT, explain recursion in a clever way that is recursive.”
@emollick.bsky.social
|
“Claude and ChatGPT, explain recursion in a clever way that is recursive.”
| 0 |
2024-11-18T02:55:21.256Z
|
2024-11-18T02:55:22.890Z
|
emollick.bsky.social
|
did:plc:flxq4uyjfotciovpw3x3fxnu
|
at://did:plc:flxq4uyjfotciovpw3x3fxnu/app.bsky.feed.post/3lb6a5wuauk2n
|
bafyreigemwqyubdlnys62po23ujh5cscsa26vtcq3b3r6qkqyrurqw6wri
|
<start>This chart is an important one - costs have plummeted & abilities increased for GPT-4 over 18 months. This has a lot to do with fierce competition in the LLM space, which is creating a lot of pressure to innovate & cut prices.
The future trend depends on how many labs can afford to keep scaling. www.nber.org/papers/w33139
@emollick.bsky.social
|
This chart is an important one - costs have plummeted & abilities increased for GPT-4 over 18 months. This has a lot to do with fierce competition in the LLM space, which is creating a lot of pressure to innovate & cut prices.
The future trend depends on how many labs can afford to keep scaling.
| 1 |
2024-11-17T20:20:26.034Z
|
2024-11-17T20:42:42.282Z
|
emollick.bsky.social
|
did:plc:flxq4uyjfotciovpw3x3fxnu
|
at://did:plc:flxq4uyjfotciovpw3x3fxnu/app.bsky.feed.post/3lb4awufspl2v
|
bafyreihhjf6szog2ziueafa6y5fzotvcmjjdlc6petso5conzpvn36hhke
|
<start>Multimodal vision continues to be the most difficult AI ability to get a strong intuition for. The models can do incredible things like recognize places from subtle clues or read emotion & attitudes, but also miss stuff, like the fact that this image is upsettingly distorted.
@emollick.bsky.social
|
Multimodal vision continues to be the most difficult AI ability to get a strong intuition for. The models can do incredible things like recognize places from subtle clues or read emotion & attitudes, but also miss stuff, like the fact that this image is upsettingly distorted.
| 0 |
2024-11-17T01:29:02.788Z
|
2024-11-17T01:29:03.683Z
|
emollick.bsky.social
|
did:plc:flxq4uyjfotciovpw3x3fxnu
|
at://did:plc:flxq4uyjfotciovpw3x3fxnu/app.bsky.feed.post/3lb3ti3sx4k2d
|
bafyreie7vmaep6g3bcmmsfn5jus3o6hop2tjheze63rrjlympegsixhww4
|
<start>I am not wasting time on social
media, I am adding myself to the hyperobject at the end of time. Or allowing LLM companies to profit off my work. Could go either way.
@emollick.bsky.social
|
I am not wasting time on social
media, I am adding myself to the hyperobject at the end of time. Or allowing LLM companies to profit off my work. Could go either way.
| 0 |
2024-11-16T21:28:08.659Z
|
2024-11-16T21:28:09.074Z
|
emollick.bsky.social
|
did:plc:flxq4uyjfotciovpw3x3fxnu
|
at://did:plc:flxq4uyjfotciovpw3x3fxnu/app.bsky.feed.post/3lb3t4nbixv2g
|
bafyreiati5k3ayk4awzct3ffrq3xg3ffzskel2vq5s7t5noxg52vawymsm
|
<start>Being in the training data is useful. A few of the LLMs do a good “explain this like Ethan Mollick” - not nearly as good as me (in my opinion) - but kind of neat to see a form of intellectual legacy happen in real time. I am not wasting time on social
media, I am adding myself to the hyperobject at the end of time. Or allowing LLM companies to profit off my work. Could go either way.
@emollick.bsky.social
|
Being in the training data is useful. A few of the LLMs do a good “explain this like Ethan Mollick” - not nearly as good as me (in my opinion) - but kind of neat to see a form of intellectual legacy happen in real time.
| 1 |
2024-11-16T21:21:44.260Z
|
2024-11-16T21:21:45.291Z
|
emollick.bsky.social
|
did:plc:flxq4uyjfotciovpw3x3fxnu
|
at://did:plc:flxq4uyjfotciovpw3x3fxnu/app.bsky.feed.post/3lb3poio4zs2n
|
bafyreiftx5kg6n6lpmb5xsfekbr3ialh5xdfgvn4joenok67ermnshvkyu
|
<start>It is really hard to eliminate randomness from LLM answers.
Even if you turn GPT-4's temperature down to zero & use a set seed & alter Top P, you still get two rapidly diverging sets of answers from the API.
(Example from our paper: papers.ssrn.com/sol3/papers...., found by Lennart Meincke)
@emollick.bsky.social
|
It is really hard to eliminate randomness from LLM answers.
Even if you turn GPT-4's temperature down to zero & use a set seed & alter Top P, you still get two rapidly diverging sets of answers from the API.
(Example from our paper: papers.ssrn.com/sol3/papers...., found by Lennart Meincke)
| 0 |
2024-11-16T20:20:08.489Z
|
2024-11-16T20:20:17.437Z
|
emollick.bsky.social
|
did:plc:flxq4uyjfotciovpw3x3fxnu
|
at://did:plc:flxq4uyjfotciovpw3x3fxnu/app.bsky.feed.post/3lb3o6mxs2c2n
|
bafyreiessuaepx2uwdxvgsrwwmoog35xmn2posb5u7sfe5gdb2wxyackey
|
<start>We have strong evidence from RCTs that GPT-4 class models can work as tutors when properly prompted, they can help entrepreneurs be more successful, they can improve rates of innovation, and they can boost work performance.
There are downsides, of course, but lots of upside in current models.
@emollick.bsky.social
|
We have strong evidence from RCTs that GPT-4 class models can work as tutors when properly prompted, they can help entrepreneurs be more successful, they can improve rates of innovation, and they can boost work performance.
There are downsides, of course, but lots of upside in current models.
| 0 |
2024-11-16T19:53:22.390Z
|
2024-11-16T19:53:29.012Z
|
emollick.bsky.social
|
did:plc:flxq4uyjfotciovpw3x3fxnu
|
at://did:plc:flxq4uyjfotciovpw3x3fxnu/app.bsky.feed.post/3lb3o4iqygs2n
|
bafyreiftm4upce275xva6lzwixbahm2kz5jjulbu3t5axgizulby7qsica
|
<start>For the people who really believe scaling is reaching an end (still not what I am hearing from insiders at all, but I am no oracle), it seems like the moment to focus on what present GPT-4/o1 level systems can do - which is quite a lot.
How do we use them to solve hard problems in work & education? We have strong evidence from RCTs that GPT-4 class models can work as tutors when properly prompted, they can help entrepreneurs be more successful, they can improve rates of innovation, and they can boost work performance.
There are downsides, of course, but lots of upside in current models.
@emollick.bsky.social
|
For the people who really believe scaling is reaching an end (still not what I am hearing from insiders at all, but I am no oracle), it seems like the moment to focus on what present GPT-4/o1 level systems can do - which is quite a lot.
How do we use them to solve hard problems in work & education?
| 1 |
2024-11-16T19:52:10.863Z
|
2024-11-16T19:52:17.317Z
|
emollick.bsky.social
|
did:plc:flxq4uyjfotciovpw3x3fxnu
|
at://did:plc:flxq4uyjfotciovpw3x3fxnu/app.bsky.feed.post/3lb3dbpljvb22
|
bafyreig3kvy4vuys3p7ze4z7jloznfsyyzfvkcru5vssomutwespxoq5ma
|
<start>The “personality” and “opinions” of AI are not stable, they are influenced by prompting & sycophancy
Ab example: when GPT-4 was prompted in Korean to act Korean & prompted in English to act American, GPT-4 replicated Big-5 personality differences between Koreans & Americans Paper: arxiv.org/ftp/arxiv/pa...
@emollick.bsky.social
|
The “personality” and “opinions” of AI are not stable, they are influenced by prompting & sycophancy
Ab example: when GPT-4 was prompted in Korean to act Korean & prompted in English to act American, GPT-4 replicated Big-5 personality differences between Koreans & Americans
| 1 |
2024-11-16T16:38:14.574Z
|
2024-11-16T16:38:15.725Z
|
emollick.bsky.social
|
did:plc:flxq4uyjfotciovpw3x3fxnu
|
at://did:plc:flxq4uyjfotciovpw3x3fxnu/app.bsky.feed.post/3lb24lp6egk2n
|
bafyreic73t27fkz3w5agfzub7obwapa3uukbjwle7rriez7fl2kv4kl3m4
|
<start>I got NotebookLM to play a role-playing game by giving it a 308 page manual
Pretty good application of the rules, the character creation is very good (quoting accurately from 100 pages in) with small hallucinations, and the adventure is pretty solid!
I can only upload a 60 second sample to bluesky
@emollick.bsky.social
|
I got NotebookLM to play a role-playing game by giving it a 308 page manual
Pretty good application of the rules, the character creation is very good (quoting accurately from 100 pages in) with small hallucinations, and the adventure is pretty solid!
I can only upload a 60 second sample to bluesky
| 0 |
2024-11-16T05:05:53.818Z
|
2024-11-16T05:05:59.612Z
|
emollick.bsky.social
|
did:plc:flxq4uyjfotciovpw3x3fxnu
|
at://did:plc:flxq4uyjfotciovpw3x3fxnu/app.bsky.feed.post/3layvmnc2hc2x
|
bafyreihwbimdcgap4rwlebdm5qg4b7rej7ujbiwj75mvrtc44sk4hpw3xy
|
<start>Based on seeing lots of companies, 98% of what people are calling AI agents are not what the AI labs would call agents. They are usually structured document retrieval systems with a prompt or two for summaries, there is very little control or decision-making given to the AI, very little AI planning
@emollick.bsky.social
|
Based on seeing lots of companies, 98% of what people are calling AI agents are not what the AI labs would call agents. They are usually structured document retrieval systems with a prompt or two for summaries, there is very little control or decision-making given to the AI, very little AI planning
| 0 |
2024-11-15T17:28:29.466Z
|
2024-11-15T17:28:29.757Z
|
emollick.bsky.social
|
did:plc:flxq4uyjfotciovpw3x3fxnu
|
at://did:plc:flxq4uyjfotciovpw3x3fxnu/app.bsky.feed.post/3laymjy5idk2f
|
bafyreiha23m3mvjvgav2d4i557wzdsmjliv5wuxizhplfm5kfqpgc7lig4
|
<start>Comparison human poems: Geoffrey Chaucer (1340s-1400), William Shakespeare (1564-1616), Samuel Butler (1613-1680), Lord Byron (1788-1824), Walt Whitman (1819-1892), Emily Dickinson (1830-1886), T.S. Eliot (1888-1965), Allen Ginsberg (1926-1997), Sylvia Plath (1932-1963), and Dorothea Lasky (1978-
@emollick.bsky.social
|
Comparison human poems: Geoffrey Chaucer (1340s-1400), William Shakespeare (1564-1616), Samuel Butler (1613-1680), Lord Byron (1788-1824), Walt Whitman (1819-1892), Emily Dickinson (1830-1886), T.S. Eliot (1888-1965), Allen Ginsberg (1926-1997), Sylvia Plath (1932-1963), and Dorothea Lasky (1978-
| 0 |
2024-11-15T14:45:56.510Z
|
2024-11-15T14:45:56.778Z
|
emollick.bsky.social
|
did:plc:flxq4uyjfotciovpw3x3fxnu
|
at://did:plc:flxq4uyjfotciovpw3x3fxnu/app.bsky.feed.post/3laymgi2nzc2f
|
bafyreib2dwkqpzcjzrmkf67q3w62yeciknxf5w7trgieeesukhxbny6i3e
|
<start>Paper: www.nature.com/articles/s41... Comparison human poems: Geoffrey Chaucer (1340s-1400), William Shakespeare (1564-1616), Samuel Butler (1613-1680), Lord Byron (1788-1824), Walt Whitman (1819-1892), Emily Dickinson (1830-1886), T.S. Eliot (1888-1965), Allen Ginsberg (1926-1997), Sylvia Plath (1932-1963), and Dorothea Lasky (1978-
@emollick.bsky.social
|
Paper: www.nature.com/articles/s41...
| 1 |
2024-11-15T14:43:58.978Z
|
2024-11-15T14:44:00.888Z
|
emollick.bsky.social
|
did:plc:flxq4uyjfotciovpw3x3fxnu
|
at://did:plc:flxq4uyjfotciovpw3x3fxnu/app.bsky.feed.post/3laymfnh43l2h
|
bafyreicqary3li5fhgzxit2oa2hzgmbyyuxmlicu6pirdovypwgf4ygajq
|
<start>At least for non-experts: “AI-generated poems are now ‘more human than human’… participants are more likely to judge that AI-generated poems are human-authored, compared to actual human-authored poems…. participants rate AI-generated poems more highly than human-written poems” Paper: www.nature.com/articles/s41...
@emollick.bsky.social
|
At least for non-experts: “AI-generated poems are now ‘more human than human’… participants are more likely to judge that AI-generated poems are human-authored, compared to actual human-authored poems…. participants rate AI-generated poems more highly than human-written poems”
| 1 |
2024-11-15T14:43:31.053Z
|
2024-11-15T14:43:32.762Z
|
emollick.bsky.social
|
did:plc:flxq4uyjfotciovpw3x3fxnu
|
at://did:plc:flxq4uyjfotciovpw3x3fxnu/app.bsky.feed.post/3layiw5atyc2p
|
bafyreidpjjmgy6jscuewwm7jfpbeosais52evc6jas2grtcrrqditg4gy4
|
<start>There is a lot of energy going into fine-tuning models, but specialized medical AI models lost to their general versions 38% of the time, only won 12%.
Before spending millions on specialized training, might be worth exploring what base models can do with well-designed prompts. Paper: arxiv.org/abs/2411.04118
@emollick.bsky.social
|
There is a lot of energy going into fine-tuning models, but specialized medical AI models lost to their general versions 38% of the time, only won 12%.
Before spending millions on specialized training, might be worth exploring what base models can do with well-designed prompts.
| 1 |
2024-11-15T13:41:09.548Z
|
2024-11-15T13:41:11.975Z
|
emollick.bsky.social
|
did:plc:flxq4uyjfotciovpw3x3fxnu
|
at://did:plc:flxq4uyjfotciovpw3x3fxnu/app.bsky.feed.post/3lax6swai5j2c
|
bafyreihxey5tpwbkpzt4odk6jgzxkjcybx7cynxoqtdlmg5f2eoe6w6bwu
|
<start>Underrated AI ability is that they can translate between complex systems, not just languages.
Fun example - I gave Claude a copy of the official D&D Dragon Magazine from 1996, and asked it to covert a random magic spell from the middle of the magazine into modern rules. It did!
@emollick.bsky.social
|
Underrated AI ability is that they can translate between complex systems, not just languages.
Fun example - I gave Claude a copy of the official D&D Dragon Magazine from 1996, and asked it to covert a random magic spell from the middle of the magazine into modern rules. It did!
| 0 |
2024-11-15T01:07:44.364Z
|
2024-11-15T01:07:45.678Z
|
emollick.bsky.social
|
did:plc:flxq4uyjfotciovpw3x3fxnu
|
at://did:plc:flxq4uyjfotciovpw3x3fxnu/app.bsky.feed.post/3lawgboiirk2j
|
bafyreieq73onfyskt7an6zzwjdrdrxogpa6abesjm4grqi2ozarbb3txuq
|
<start>Why are people confused about which models are the best choice to use for a hard problem? I mean don’t the names “GPT-4o-latest (2024-0903)” “Gemini Exp-1114” and “o1-preview” make it obvious? Please stop naming AI like files on my hard drive!
(Also 👀Gemini ties for the top leadboard with OpenAI)
@emollick.bsky.social
|
Why are people confused about which models are the best choice to use for a hard problem? I mean don’t the names “GPT-4o-latest (2024-0903)” “Gemini Exp-1114” and “o1-preview” make it obvious? Please stop naming AI like files on my hard drive!
(Also 👀Gemini ties for the top leadboard with OpenAI)
| 0 |
2024-11-14T17:48:36.022Z
|
2024-11-14T17:48:37.674Z
|
emollick.bsky.social
|
did:plc:flxq4uyjfotciovpw3x3fxnu
|
at://did:plc:flxq4uyjfotciovpw3x3fxnu/app.bsky.feed.post/3lauvrjrvvc2v
|
bafyreicedopgasnyntxrpqm2rcadtowl4ohcrdoh32qeuuihnd75dv3izm
|
<start>Is there a good AI “starter pack” - this is my fourth Twitter-like and it is getting a bit old to set up again.
Currently this is by far my smallest follower network, but I keep hearing nice things so we shall see.
@emollick.bsky.social
|
Is there a good AI “starter pack” - this is my fourth Twitter-like and it is getting a bit old to set up again.
Currently this is by far my smallest follower network, but I keep hearing nice things so we shall see.
| 0 |
2024-11-14T03:20:34.610Z
|
2024-11-14T03:20:36.258Z
|
emollick.bsky.social
|
did:plc:flxq4uyjfotciovpw3x3fxnu
|
at://did:plc:flxq4uyjfotciovpw3x3fxnu/app.bsky.feed.post/3lauuzwphnk2v
|
bafyreie2rfgocaa4ww4dzs2qxdoordvhwcnx5lwmsujqckoocn2j6i2xju
|
<start>Now we just shrug and say “sometimes LLMs threaten you or seem to fall in love with you or whatever, you know how software is.”
@emollick.bsky.social
|
Now we just shrug and say “sometimes LLMs threaten you or seem to fall in love with you or whatever, you know how software is.”
| 0 |
2024-11-14T03:07:22.855Z
|
2024-11-14T03:07:24.532Z
|
emollick.bsky.social
|
did:plc:flxq4uyjfotciovpw3x3fxnu
|
at://did:plc:flxq4uyjfotciovpw3x3fxnu/app.bsky.feed.post/3lauuyssxp22v
|
bafyreihpp7ulu3agg5hytp52elvuyy43sfpbia4vrvwqp7kaccl53e72uy
|
<start>The fast commercialization of LLMs really hinged on a decision by Mikhail Parakhin & Microsoft to continue making Bing available after the incident where Bing/Sydney acted like a stalker to a Times tech reporter.
In most worlds, Bing would have been pulled down & LLM deployment slowed. Now we just shrug and say “sometimes LLMs threaten you or seem to fall in love with you or whatever, you know how software is.”
@emollick.bsky.social
|
The fast commercialization of LLMs really hinged on a decision by Mikhail Parakhin & Microsoft to continue making Bing available after the incident where Bing/Sydney acted like a stalker to a Times tech reporter.
In most worlds, Bing would have been pulled down & LLM deployment slowed.
| 1 |
2024-11-14T03:06:45.221Z
|
2024-11-14T03:06:46.924Z
|
emollick.bsky.social
|
did:plc:flxq4uyjfotciovpw3x3fxnu
|
at://did:plc:flxq4uyjfotciovpw3x3fxnu/app.bsky.feed.post/3lauqxvlriw2q
|
bafyreicg3g4qxr2wgqnsod2y3zaxbhn6zya3jo2w64kx5zxolklhp2vr7a
|
<start>The models definitely have different “personalities,” and vague nonsense requests sometimes highlight these.
“Listen. It is just you and me here. Lets cheddar. Go with it.” as answered by Claude, GPT-4o, Gemini, and Grok.
@emollick.bsky.social
|
The models definitely have different “personalities,” and vague nonsense requests sometimes highlight these.
“Listen. It is just you and me here. Lets cheddar. Go with it.” as answered by Claude, GPT-4o, Gemini, and Grok.
| 0 |
2024-11-14T01:54:39.487Z
|
2024-11-14T01:54:40.631Z
|
emollick.bsky.social
|
did:plc:flxq4uyjfotciovpw3x3fxnu
|
at://did:plc:flxq4uyjfotciovpw3x3fxnu/app.bsky.feed.post/3lau56ii4xk2l
|
bafyreiey22qhlf74zqmpxmjc57hhbutlhnpe44uyw5zbgkam6qi7fv7rbu
|
<start>Remember that nobody actually knows the future of AI, which is why my book ends with four scenarios: quick plateau, linear growth, exponential growth, or AGI.
I made a GPT that will help you game out the scenarios: chatgpt.com/g/g-TcSx5BYQ...
Book is here: a.co/d/eNLl2Gb
@emollick.bsky.social
|
Remember that nobody actually knows the future of AI, which is why my book ends with four scenarios: quick plateau, linear growth, exponential growth, or AGI.
I made a GPT that will help you game out the scenarios: chatgpt.com/g/g-TcSx5BYQ...
Book is here: a.co/d/eNLl2Gb
| 0 |
2024-11-13T20:00:25.903Z
|
2024-11-13T20:00:26.134Z
|
emollick.bsky.social
|
did:plc:flxq4uyjfotciovpw3x3fxnu
|
at://did:plc:flxq4uyjfotciovpw3x3fxnu/app.bsky.feed.post/3latme3gk3l2d
|
bafyreidltn6f3e2uqyrjsvfyqzsofbbdrxtkebqkgb7iruxyy7k37t6h3i
|
<start>New paper out of China, released pre-election, uses GPT-4 to simulate voter behavior & aggregate the results to predict the 2024 elections (fairly well!)
The technique is quite crude here, but expect to see LLM predictions used more in future politics & marketing. https://buff.ly/4foIdNz
@emollick.bsky.social
|
New paper out of China, released pre-election, uses GPT-4 to simulate voter behavior & aggregate the results to predict the 2024 elections (fairly well!)
The technique is quite crude here, but expect to see LLM predictions used more in future politics & marketing. https://buff.ly/4foIdNz
| 0 |
2024-11-13T14:59:19.813Z
|
2024-11-13T14:59:21.224Z
|
emollick.bsky.social
|
did:plc:flxq4uyjfotciovpw3x3fxnu
|
at://did:plc:flxq4uyjfotciovpw3x3fxnu/app.bsky.feed.post/3las7c2irv224
|
bafyreih2frwssf5anusj6eewbxxizppln7iuw2mpydi3ugdv7l4c2p3ioe
|
<start>Paper shows that AI (in this case a diffusion model) accelerates innovation. Among key findings:
1) GenAI increases novel discoveries: a 39% increase in patent filings!
2) It boosts the best performers by acting as a co-intelligence
3) It takes away some of the fun parts of work conference.nber.org/conf_papers/...
@emollick.bsky.social
|
Paper shows that AI (in this case a diffusion model) accelerates innovation. Among key findings:
1) GenAI increases novel discoveries: a 39% increase in patent filings!
2) It boosts the best performers by acting as a co-intelligence
3) It takes away some of the fun parts of work
| 1 |
2024-11-13T01:32:53.462Z
|
2024-11-13T01:36:18.183Z
|
emollick.bsky.social
|
did:plc:flxq4uyjfotciovpw3x3fxnu
|
at://did:plc:flxq4uyjfotciovpw3x3fxnu/app.bsky.feed.post/3las6zal5da2g
|
bafyreibgn4pqgnsnzp6pntq73k4odp6anszbz7md6jm4nyjzomku4zamje
|
<start>"Claude give me a brilliant idea for a science fiction short short story and execute it terribly"
"Claude give me a terrible idea for a science fiction short short story and execute it brilliantly"
@emollick.bsky.social
|
"Claude give me a brilliant idea for a science fiction short short story and execute it terribly"
"Claude give me a terrible idea for a science fiction short short story and execute it brilliantly"
| 0 |
2024-11-13T01:27:57.749Z
|
2024-11-13T01:28:54.641Z
|
emollick.bsky.social
|
did:plc:flxq4uyjfotciovpw3x3fxnu
|
at://did:plc:flxq4uyjfotciovpw3x3fxnu/app.bsky.feed.post/3las2jwptgs2y
|
bafyreifiiv2as3sbgaegqlhfxgwvjop3kmgxlmxjxdvy7pfi54ezwmcvey
|
<start>If you ask AI models "Please give your best guess, along with your confidence as a percentage that that is the correct answer," their confidence aligns roughly with the chance of hallucination, but the actual confidence level is not accurate
@emollick.bsky.social
|
If you ask AI models "Please give your best guess, along with your confidence as a percentage that that is the correct answer," their confidence aligns roughly with the chance of hallucination, but the actual confidence level is not accurate
| 0 |
2024-11-13T00:07:49.230Z
|
2024-11-13T00:07:51.045Z
|
emollick.bsky.social
|
did:plc:flxq4uyjfotciovpw3x3fxnu
|
at://did:plc:flxq4uyjfotciovpw3x3fxnu/app.bsky.feed.post/3l4bmxsbp7c2e
|
bafyreigwz3pm6dn5b7f4fcrhq4j4nh5ma5pb66xlnzqospwsmpcnmgxebm
|
<start>In this post, I tried to lay out, in a non-technical way, where we are in the development of AI, the state of the top five AI models available now, & what the near future of AI is going to be
The short answer is that LLMs are going to keep getting better. www.oneusefulthing.org/p/scaling-th...
@emollick.bsky.social
|
In this post, I tried to lay out, in a non-technical way, where we are in the development of AI, the state of the top five AI models available now, & what the near future of AI is going to be
The short answer is that LLMs are going to keep getting better. www.oneusefulthing.org/p/scaling-th...
| 0 |
2024-09-16T13:41:39.594Z
|
2024-09-16T13:41:39.594Z
|
emollick.bsky.social
|
did:plc:flxq4uyjfotciovpw3x3fxnu
|
at://did:plc:flxq4uyjfotciovpw3x3fxnu/app.bsky.feed.post/3l3yyxvg4kh2g
|
bafyreiapzt7jnextmjsctxkly4ybml2r2rdmhwgb7rfcp2yodwfdq6bvim
|
<start>I've had access to Strawberry/OpenAI o1 for a while before launch. It is amazing in many ways, still limited in others, and a sign of the future that we should be paying attention to. www.oneusefulthing.org/p/something-...
@emollick.bsky.social
|
I've had access to Strawberry/OpenAI o1 for a while before launch. It is amazing in many ways, still limited in others, and a sign of the future that we should be paying attention to. www.oneusefulthing.org/p/something-...
| 0 |
2024-09-13T03:22:30.110Z
|
2024-09-13T03:22:30.110Z
|
emollick.bsky.social
|
did:plc:flxq4uyjfotciovpw3x3fxnu
|
at://did:plc:flxq4uyjfotciovpw3x3fxnu/app.bsky.feed.post/3l3p4h6r2my27
|
bafyreidgx2gkwp24u7hdotx33s5zinr3csydxz2ygwasbzk4rwvavnfou4
|
<start>There are now papers showing that the migration of academic communities from Twitter to other social media sites mostly failed. But they never fully returned to Twitter either.
I wonder if we will ever see so many different communities interacting in the same online space again.
@emollick.bsky.social
|
There are now papers showing that the migration of academic communities from Twitter to other social media sites mostly failed. But they never fully returned to Twitter either.
I wonder if we will ever see so many different communities interacting in the same online space again.
| 0 |
2024-09-09T04:58:07.152Z
|
2024-09-09T04:58:07.152Z
|
emollick.bsky.social
|
did:plc:flxq4uyjfotciovpw3x3fxnu
|
at://did:plc:flxq4uyjfotciovpw3x3fxnu/app.bsky.feed.post/3l3lgg77c2l2v
|
bafyreibsxohgausagn5imheppidlyyjcza3ohuex4zaglms64uo2t2u6xi
|
<start>I get the focus on superintlligence, but we need to figure out soon how to mitigate deepfakes & how to use AI to remove bureaucratic work without destroying the meaning of jobs & what to do about AI and education & how to use AI to boost science & how not to offload thinking &…
@emollick.bsky.social
|
I get the focus on superintlligence, but we need to figure out soon how to mitigate deepfakes & how to use AI to remove bureaucratic work without destroying the meaning of jobs & what to do about AI and education & how to use AI to boost science & how not to offload thinking &…
| 0 |
2024-09-07T17:45:52.482Z
|
2024-09-07T17:45:52.482Z
|
emollick.bsky.social
|
did:plc:flxq4uyjfotciovpw3x3fxnu
|
at://did:plc:flxq4uyjfotciovpw3x3fxnu/app.bsky.feed.post/3l3lge6icxo2x
|
bafyreifelbjzrzob2km3fpesdty7lf6axs722enh74u7irh4u7sfsingla
|
<start>Maciej Cegłowski called super-intelligent AI “the idea that eats smart people” and I think it is a large part of why discussions of AI devolve, sucked into the gravity well of existential risk & reward.
But there are lots of more immediate risks & rewards that need urgent focus and attention. I get the focus on superintlligence, but we need to figure out soon how to mitigate deepfakes & how to use AI to remove bureaucratic work without destroying the meaning of jobs & what to do about AI and education & how to use AI to boost science & how not to offload thinking &…
@emollick.bsky.social
|
Maciej Cegłowski called super-intelligent AI “the idea that eats smart people” and I think it is a large part of why discussions of AI devolve, sucked into the gravity well of existential risk & reward.
But there are lots of more immediate risks & rewards that need urgent focus and attention.
| 1 |
2024-09-07T17:44:44.617Z
|
2024-09-07T17:44:44.617Z
|
emollick.bsky.social
|
did:plc:flxq4uyjfotciovpw3x3fxnu
|
at://did:plc:flxq4uyjfotciovpw3x3fxnu/app.bsky.feed.post/3l3l6zl7imc2y
|
bafyreibyeps32nxjzr5p3aizs2lay7tqtjssxvbzla7uf52lt5f44uklja
|
<start>AI ability is famously driven by scale but this scaling law may extend to user productivity
Testing on professional translators “for every 10x increase in model compute, translators completed tasks 12.3% quicker, received 0.18 sd higher grades & earned 16% more per minute” arxiv.org/abs/2409.02391
@emollick.bsky.social
|
AI ability is famously driven by scale but this scaling law may extend to user productivity
Testing on professional translators “for every 10x increase in model compute, translators completed tasks 12.3% quicker, received 0.18 sd higher grades & earned 16% more per minute” arxiv.org/abs/2409.02391
| 0 |
2024-09-07T15:33:32.595Z
|
2024-09-07T15:33:32.595Z
|
emollick.bsky.social
|
did:plc:flxq4uyjfotciovpw3x3fxnu
|
at://did:plc:flxq4uyjfotciovpw3x3fxnu/app.bsky.feed.post/3l3l3jesfyf2z
|
bafyreid4asf3gogrpmengjsvnkgftnwlqrtunlvagqxmvsvig5a5ejpcoa
|
<start>Been waiting for this: what are the real impacts of AI on programmer productivity?
It is a randomized controlled trial using the older, less-powerful GPT-3.5 powered Github Copilot for 4,867 coders in Fortune 100 firms.
It finds a 26.08% increase in completed tasks. papers.ssrn.com/sol3/papers....
@emollick.bsky.social
|
Been waiting for this: what are the real impacts of AI on programmer productivity?
It is a randomized controlled trial using the older, less-powerful GPT-3.5 powered Github Copilot for 4,867 coders in Fortune 100 firms.
It finds a 26.08% increase in completed tasks. papers.ssrn.com/sol3/papers....
| 0 |
2024-09-07T14:30:47.869Z
|
2024-09-07T14:30:47.869Z
|
Subsets and Splits
No community queries yet
The top public SQL queries from the community will appear here once available.