Ethan Mollick (@emollick) header
Ethan Mollick (@emollick) profile picture

Professor @Wharton studying AI. New book, Co-Existence, is out now: https://teapawt.org/t.co/hsDYf1xU9K My Substack: https://teapawt.org/t.co/UIBhxu4bgq

Philadelphia, PA
Joined May 2009
Photo by Ethan Mollick (@emollick)Photo by Ethan Mollick (@emollick)Photo by Ethan Mollick (@emollick)Photo by Ethan Mollick (@emollick)Photo by Ethan Mollick (@emollick)Photo by Ethan Mollick (@emollick)Photo by Ethan Mollick (@emollick)Photo by Ethan Mollick (@emollick)Photo by Ethan Mollick (@emollick)Photo by Ethan Mollick (@emollick)
My new book, Co-Existence, is out today. It is a guide to how to live & work in a world where AI is sometimes, but not always, better than you. (And, for the release, I worked with Opus 5.5 to make a quite interesting experience for the book's website: co-existence.ai/)
210
86
38
1,064
154,956
I’d note that OpenAI has always been great about not influencing early testers & has encouraged candid feedback. The mixed initial 6.2 reactions from testers didn’t blindside them. Some model point releases are just weirder, I wouldn’t overindex on it as a sign of the future.
14
3
1
66
5,311
I am sure there are many use cases where 6.2 will be the best model. It can also be good to have models that differ from each other in strengths and weaknesses. And in some short period of time there will be another version of Astra, and the existing models will be tweaked.
4
8
1,784
All of this focus on individual releases also misses the broader point with much bigger implications for all of us: rapid, continuing acceleration of model development is very obviously happening and is the thing to actually pay attention to.
3
1,167
Google, OpenAI & Anthropic are at the stage where, unless you frequently switch between them or actively test them, you are probably very happy with the series of releases in the past two months. Even the model releases that are not as strong represent large capability gains
223
10
15
657
30,620
Most people don’t realize how much sickness we all lived with before modern medicine. The measles vaccination alone is responsible for half a percent of all US income this year, $80 billion. That is just from the reduced disease burden, it doesn’t include reduced medical costs
13
68
2
442
23,676
All of that aside, the fact that an AI can make a dynamically lit explorable 3D scene set in a realistic space elevator with realistic physics where every single object was coded by "hand" remains crazy. If you have a decent computer, it is worth trying the web version: upwell-tetherline.netlify.ap… Or download the full Windows install from itch. io: emollick.itch.io/upwell
I’ve been using GPT-6.2 Astra prior to launch & I found it very uneven & weird. On one hand, it was harder to steer & work with than other models. On the other, when given an ambitious project (make a space elevator ride from scratch with Blender & Unity) it could deliver well
40
7
1
169
24,407
All the switches work too!
1
12
8,196
I know a lot of people are discovering this about GPT-6.2, but the GPT models have always been obsessed with bells and drowned clock towers. Always.
Forget goblins, things that GPT-5.5 really likes in its fiction: lighthouses, the ocean, maps, bells, clock towers with bells that ring impossible times, Mira Vale, resonances and echoes (Claude and Gemini love them too), secret third things (not night/day, not high/low)...
47
4
3
289
26,150
They all have the aesthetics of Ken Levine.
19
8,347
A rush of patches to address security concerns is exactly the most plausible answer for recent outages, as well as the official reason. CIOs I talked to have been exhausted trying to keep up with security patch demands & balacing that with normal operations.
NEWS: Wall Street execs and NSA have told Congress that outages were caused by patching vulnerabilities identified by AI models, sources say AI did not cause outages but found so many issues that the rush to fix them before an attack overwhelmed systems washingtonsun.com/technology…
70
69
6
681
39,539
Anyhow, I do not think information will stop the conspiracy theories. (And yes, when Mythos-class unguardrailed models are available, that will also lead to a lot of issues)
1
35
8,127
Folks who are very close to AI news on X tend to extrapolate too much from each new model. Anthropic went through a rough patch a few months ago with Opus releases before 5.5, and goodness knows that Google had a long dry stretch. 6.2 is not terrific but that doesn’t tell us much
Model personality matters in a way that benchmarks can't capture. Opus models went through a rough patch from around 4.7 to 5 where they just didn't feel "Claude-y" anymore, more like an watered-down Fable (hmmm, teacher models?). Opus 5.5 feels like working with ol' Claude again
68
9
5
489
31,134
Ha. When I had Astra 6.2 make a puzzle spy game using any setting it wanted, it set the game in "Bellwether Station." I thought it was an odd choice but didn't realize how deep a pattern this was. What a weird attractor for the model.
I was saving this for my video but I'll spoil it now. Ben and I both made new levels in games using 6.2 Astra during the early testing. He did Majora's Mask, I did Outer Wilds. We both prompted ourselves, but it was along the lines of "make a creative new level for [game]" We both got water levels with music themes, specifically around bells. Both were named "Bellwether". This model regresses to the mean like nothing I've ever seen. Go ask it to make a new level for a game you love, lmk if it chooses "bellwether" as the name.
71
30
15
1,260
98,591
An illustration of both the power and issues with Astra 6.2 in a way that might be accessible: I asked it to create a game in Blender and Unity where you are the pilot of a Coast Guard cutter who travelled back in time to the War of Independence to stop the British invasion of New York. It made incredible 3D models and a working game with no errors - its even fun to play! But Astra got obsessed with random ideas that didn't make a lot of sense and built them out without asking along with following my instructions: you could put down false buoys to lure the ship in the wrong direction (like the British would understand a modern buoy?) or you could persuade the British ships to go the wrong direction by talking to them, or that you would spend your time rescuing sailors from other parts of the harbor. It didn't even initially have a shooting mechanism until I pointed it out. This is all kind of minor stuff individually, but it was frustrating to have the AI constantly do its own things without telling me, given its instincts were often off, and then have to redirect it constantly. Other models, including GPT-6 Asta, didn't require the same struggle to control and operate. And these same issues show up in academic research and other more serious use cases.
39
10
5
435
35,179
The reviews of the early testers like myself of GPT 6.2 Astra is a good vindication of the continuing value of vibe benchmarks in addition to numeric ones. We lack good measures of personality or perseverance or creativity or writing clarity or taste. But these things matter
GPT-6.2 Astra scores 174 on the Epoch Capabilities Index (ECI), on par with Fable 5.2 and joining it above what the pre-Astra trend for reasoning models would have predicted. This is based on testing we did prior to launch using only our own benchmarks. The ECI is continually updated as more benchmark results become available, and so these values may change in the future.
68
10
6
544
38,865
I’ve been using GPT-6.2 Astra prior to launch & I found it very uneven & weird. On one hand, it was harder to steer & work with than other models. On the other, when given an ambitious project (make a space elevator ride from scratch with Blender & Unity) it could deliver well
143
58
33
1,304
123,475
You can play the world in your browser here, it even simulates zero g physics. upwell-tetherline.netlify.ap… Though this impressive (& it is), I don’t find myself reaching for the new 6.2 models. They really are hard to work with & take more guidance and pushing to get work done.
2
3
1
124
20,041
This is the first time since GPT 4.5 that a frontier release from OpenAi is not a clear step forward. I’m willing to bet that 6.2 is amazing at some types of complex work, but in my experience & that of others I know, it was often frustrating to use.
10
9
5
207
18,642
Ethan Mollick retweeted
I dug into this a bit. It looks like we're at highs, but not unprecedented highs for major systems failures. We're seeing alarming headlines every day, but a lot of it is recency bias.
Is it possible we're hearing about this stuff more because people are more jumpy about cyberattacks?
53
278
57
2,472
150,508
The discussion of AI in gaming is pernicious now. The idea that games are going to be the only field where humans write code unassisted is obviously wrong. But the push for that means missing a chance to guide AI usage toward approaches that empower workers & help gaming thrive.
94
11
5
342
26,559
Games have always been a very difficult industry with a high cost to its workers in terms of crunch & instability. I would love to see more positive visions about how you rebuild an industry stronger using AI, rather than defending the status quo as good and AI is going away.
9
1
64
10,745
AI is a general purpose technology has good and bad impacts, the general model should be improving the good and building on it, while mitigating the bad. Seeing AI as all one thing is a barrier.
Replying to @emollick
Honestly most players I see aren't mad about devs using AI to help write code, they're mad about AI art and low effort shovelware. Lumping it all into one fight makes it way harder to push for the good uses.
1
35
7,244