@VictorTaelin First time I've disagreed w/ one of your posts, tbh. Patience my friend. Scientific progress is more important than devs' projects (actually I've disagreed before - these models will be able to do new science)
English
Jordan Moren
118 posts
@KelseyTuoc Rich person is upset democrats are doing something; News at 11
English
We should ban partisan gerrymandering nationwide. It's bad for our country. Republicans control Congress and should introduce the bill. Democrats will support it.
English
@emollick There are several thresholds after this. Can have the same qualitative leap a dozen more times and still be able to notice the difference.
English
I have been using GPT ImageGen-2 for the past weeks
I didn't think that better image-generators would be a big deal but it turns out that there is a quality threshold I didn't expect, where you can now get text, slides, academic papers
Look at what it does with my "otter test"!
English
@npew the GPUs are 2kW not 1; closer to 2.2-2.5kW considering all adjacent systems. So more like 10-15M GPUs, which is a big difference.
English
This is is still early. 30 GW ~= 30m GPUs. Long way to go for one GPU per human.
OpenAI Newsroom@OpenAINewsroom
In January 2025, we committed to generating 10GW of compute and have already identified over 8GW of that. Now, we're planning for 30GW of compute by 2030. A milestone that scales with the rapidly accelerating demand for intelligent systems. Image generated by @ChatGPTapp Images 2.0 😉 #YouCanJustBuildThings
English
@GaryMarcus LOL oops I guess I'm not generally intelligent either
English
@PazneriaJ dude it did not, see qt. or look carefully yourself. 🙄
English
There seems to be some excitement around “ChatGPT’s powerful new image engine”, but as ever, its functional understanding of the world seems limited.
I first learned about the new system when some some smart aleck on X sent me an example of the new system trying to label a bike (an example I have considered before), with the caption “Uh oh”, apparently believing that my longstanding challenges to image generation had been solved.

It does look impressive on first inspection, better than some examples I showed here before.
But if you look closely, there are several errors, and those errors are revealing. For example, the rear center-pull (?) brake was mislabeled as the seat stay, and the big gear on the back was mislabeled as the rear brake. There was a label for a spoke that is pointing to blank space.
In many modern bikes, of course, a rear brake can be found back there, but not in this diagram. Instead this system has combined a typical position for a modern disc brake system with a diagram of an older (though still in use) caliper (or similar) system. The system doesn’t actually understand how the various parts function.
And of course there are literally hundreds of labeled bikes on the internet as a quick Google Image Search would reveal. (Which is why my usual test here has been a tandem bike, to make things a little more challenging.)
To up the degree of difficulty, I asked ChatGPT to “please draw a taller than average tandem bike, and include a bike rack and panniers”, which is not something you could readily find on the internet, and not something I used here before, and got the bike-wreck of a picture I include below.
Bike nuts would have a field day finding problems with it. (Feel free to drop your favorite error in the comments).
Suffice to say that most people don’t stuff their rear derailleur in the back wheel. And I don’t even know what to say about that “rear brake lever”, or the saddle-shaped rear handlebar, let alone the “rear brake” that is somehow part of the rear rack.
As in the first example, the lack of functional understanding is manifest.
Of course, to be fair, the average human couldn’t complete this task, either.
But anybody knowledgeable about bikes (racers, mechanics, designers etc) would immediately see numerous problems.
And honestly is anybody tall enough to ride in the front?
English
@GaryMarcus And it gave you one. My point is that you're criticizing these things based on the poorest performance available - can you at least acknowledge that it did it right in the image I shared?
Sure I'll give it a try do you mind sharing the prompt you used?
English
@PazneriaJ the system literally promised me a powerful new image engine to try.
i am not going to pay money to that unethical company but you are welcome to try my tall tandem w panniers example and post the output here
English
@ericmitchellai It needs to not rely on the user being technical and knowledgeable. If I'm asking it to do something in a stupid way and it knows a better, established, way to do the thing that I want it should bring that up to me
English
why isn't chatgpt the perfect personal AGI?
what is most disappointing about it?
what feature, model improvement, or bugfix would do the most to make it more useful in your daily life?
what is most frustrates you that chatty can't do, or can't do well enough?
English
@thsottiaux ALSO - I, and I would guess most non-technical people (which is most people) reason through conversation. ChatGPT knows all sorts of random ideas I've had over the years, Codex does not. Please integrate ChatGPT memory with Codex.
English
@HarryStebbings It's interesting to watch the normies progressively get more into AI. In five years you won't need to know any of that - the models will, LOL
English
What Role Does Not Exist Today But Will Be So Common in Five Years Time:
"500K-1M jobs will be created for agent operators.
This person will be somewhat technical. They will be deep in the AI world.
They're gonna have to understand MCPs and CLIs and they are going to have to know how to write skills.
It's going be this group of people that will know how to go into your marketing team or your legal team, or your operations team, or your life sciences research team and this is the person that is basically going to enable that function to get leverage from agents." @levie
Where is this right? Where is this wrong? @jasonlk @gregisenberg @amasad @AnjneyMidha
Harry Stebbings@HarryStebbings
The amount of hype and BS going around about enterprise AI adoption is insane. Aaron @levie is the most AI forward-thinking CEO in public markets today. But even Aaron at $1BN+ in ARR is valued at $3.3BN and getting smashed by Wall St. I sat down with Aaron to understand WTF is happening, what is real and what is fake in enterprise, WTF to do with token budgets and wrote up my notes below. (Link to full episode in comments) 1. Why Dwarkash Was Wrong and Jensen Was Right on Upgrading Systems Upgrading software is a multi-year effort, not a "magical moment" where everything can be secured overnight. The reality of enterprise security is an ongoing, endless cycle of "leapfrogging" between defensive and offensive capabilities. Founders must realize that even with access to frontier models, the implementation cycle in the real world remains the primary bottleneck. 2. Why We Will Have More Lawyers in Five Years Not Less The industry is myopic about job elimination; AI makes it easy to generate content, but it hasn’t made it easier to get that content approved by a court or a patent office. As clients inundate lawyers with AI-generated contracts and memos, the "ultimate constraint" becomes the number of qualified humans available to review and approve the output. 3. What Role Does Not Exist Today That Will Be Incredibly Common in Five Years? We are about to see the creation of 500,000 to 1 million "Agent Operators". These technical-yet-business-savvy individuals will be responsible for "care and feeding" of agents—writing skills, understanding MD files, and redesigning workflows for agents rather than people. 4. Will Massive Software Providers Simply Be Turned Into a Database That Agents Crawl Over? While the user interface may shift to chat, the value is moving to the API layer and the "business logic" embedded above the database. Systems like ERPs are more than databases; they contain decades of complex logic for supply chains and accounting that agents must interact with, not replace. 5. What Everyone Thinks About Enterprise AI Adoption That They Get Wrong The assumption that the massive gains seen in AI coding will immediately translate to all other knowledge work is a "misread". Coding has specific idiosyncrasies that don't always exist in broader knowledge work, where human collaboration and regulatory loops are more complex. 6. Where Would You Be Investing if You Were a VC Today? Despite high valuations, Levie would still be "loading up" on frontier rounds. These companies have the potential to grow much larger because the ultimate market for AI is often larger than the industry currently realizes. 7. The Budget of Tokens Will Have to Move Out of IT Spend and Into Opex Enterprise AI shouldn't be treated as a tradeoff between software licenses. Instead, token budgets will move into regular operational expenditure (OPEX), where businesses trade off a marketing campaign for a more productive, automated marketing engine. This allows AI companies to tap into a massive pool of capital beyond the traditional, capped IT budget.
English
@thsottiaux I'd like Codex to be more proactive - and, obviously, computer use on Windows (and improved!)
English
@kyliebytes at first glance I thought you were in a Planet Fitness
English
tomorrow's episode
subscribe to core memory today
Ashlee Vance@ashleevance
Tomorrow, @sama and @gdb will be on a podcast together for the first time ever. And they will give an exact date for the arrival of AGI. New models, the trial with @elonmusk, @AnthropicAI beef, Sam personal drama, the future of compute buys, @kyliebytes and I do it all. @corememory episode 67 hits on the Substack only. If you want the goods, it's time to sub up corememory.com The Core Memory pod is on all major platforms and brought to you by @brexHQ and @e1ventures because they have great taste.
English
English
@PazneriaJ @sixfootbrit @iruletheworldmo Those games wouldn't cost the 70/90 bucks today a AAA game costs, and who would pay for that? If the fad is a one prompt custom game or app it'd fade away quickly like Suno or Genie 3
English
we’re going to have to begin serious discussions around ubi or novel ways to structure the economy.
for a short period
a handful of people that can afford infinite tokens are going to pull away. we should be discussing what that might look like, how to ensure people have income and meaning in their lives
before it actually happens.
English
@venturetwins anti ai people are literally as delusional as MAGA people
English
Yesterday marked a turning point in the Iranian war. For the first time, a ship was directly attacked at sea.
Negotiations are scheduled to continue on Monday.
The global supply chain depends on this, and especially so. *Disclaimer* This is analytical reporting, not partisanship.
English
@sixfootbrit @iruletheworldmo make one game a day. make an app a day. do market research for games/apps, brainstorm and come up with a thousand ideas every day and pick the best one. think you could make a lot of money with 100 human employees you don't gotta pay?
English
@iruletheworldmo What are they going to do with the infinite tokens that allows them to pull away? Asking for a friend.
English
@iruletheworldmo "for a short period" - why lick boots? I'm serious, what's your reasoning? Why should I be enslaved when machines produce abundance? Why is that what you want? It's deeply historically uninformed - people with leisure time do the most and drive progress forward.
English