The number one worst thing they've done was when Sam tried to get the US government to regulate AI so only a handful of companies could pursue research. They wanted to protect their moat.
What's even scarier is that if they actually had the direct line of sight to AGI that they had claimed, it would have resulted in many businesses and lines of work immediately being replaced by OpenAI. They knew this and they wanted it anyway.
Thank god they failed. Our legislators had enough of a moment of clarity to take the wait and see approach.
First, when they thought they had a big lead, OpenAI argued for AI regulations (targeting regulatory capture).
Then, when lead evaporated by Anthropic and others, OpenAI argued against AI regulations (so that they can catch up, and presumably argue for regulations again).
Most regulations that have been suggested would but restrictions mostly the largest, most powerful models, so they would likely affect OpenAI/Anthropic/Google primarily before smaller upstarts would be affected.
I think you can both think there's a need for some regulation and also want to avoid regulation that effectively locks out competition. When only one company is pushing for regulation, it's a good bet that they see this as a competitive advantage.
Dude, they completely betrayed everything in their "mission". The irony in the name OpenAI for a closed, scammy, for profit company can not be lost on you.
They released a near-SOTA open-source model recently.
Their prerogative is to make money via closed-source offerings so they can afford safety work and their open-source offerings. Ilya noted this near the beginning of the company. A company can't muster the capital needed to make SOTA models giving away everything for free when their competitor is Google, a huge for-profit company.
As per your claim that they are scammy, what about them is scammy?
Their contribution to opensouurce and open research is far behind other organisations like Meta and Mistral, as welcome as their recent model release is. Former security researchers like Jan Leike commonly cite a lack of organisational focus on security as a reason for leaving.
Not sure specifically what the commenter is referring to re: scammy, but things like the Scarlett Johansson / Her voice imitation and copyright infringement come to mind for me.
Oh yeah, that reminds me. the company did research on how to train a model that manipulates the metrics, allowing them to tick the open source box with a seemingly good score, while releasing something that serves no real purpose. [1] [2]
GPT-OSS is not a near-state-of-the-art model: it is a model deliberately trained in a way that it appears great in the evaluations, but is unusable and far underperforms actual open source models like Ollama. That's scammy.
That explains why gpt-oss wasn't working anywhere near as well for me as other similarly and smaller sized models. gemma3 27b, 12b, and phi4 (14b?) all significantly outperformed it when transforming unstructured data to structured data.
Yes I am saying there is a man "sat" as it were behind every ChatGPT chat. The authors of ChatGPT basically made something closer to a turing-complete "choose-your-own adventure" book. They ensured you could choose an adventure where the reader can choose a suicide roleplay adventure, but it is up to the reader whether they want to flip to that page. If they want to flip to the page that says "suicide" then it will tell them exactly what the law is, they can only do a suicide adventure if it is a roleplaying story.
By banning chatGPT you infringe upon the speech of the authors and the client. Their "method of speech" as you put it in this case is ChatGPT.
It takes intent and effort to publish or speak. That’s not present here. None of the authors who have “contributed” to the training data of any ai bot have consented to such.
In addition, the exact method at work here - model alignment - is something that model providers specifically train models for. The raw pre training data is only the first step and doesn’t on its own produce a usable model.
So in effect the “choice” on how to respond to queries about suicide is as much influenced by OpenAIs decisions as it is by its original training data.
There are consequences to speech. If you and I are in conversation and you convince me (repeatedly, over months, eventually successfully) to commit suicide then you will be facing a wrongful death lawsuit. If you publicize books claiming known falsehoods about my person, you'll be facing a libel lawsuit. And so on.
If we argue that chatbots are considered constitutionally protected speech of their programmers or whatever, then the programmers should in turn be legally responsible. I guess this is what this lawsuit mentioned in the article is about. The principle behind this is not just about suicide but also about more mundane things like the model hallucinating falsehoods about public figures, damaging their reputation.
Just as a rule of thumb, doesn't the fact that only a few unknown vendors flag it—and all of the major vendors do not—indicate something? It would suggest a false positive, wouldn't it?
The concept of being able to (for a physical example) rip ANY page out of a book and being able to replace it using only the information on your single "magic page" is incredible. With two magic pages you can replace any two torn out pages, and the magic pages are interchangeable.
The math we have mastered is incredible. If only they could impart this wonder to children instead of rote worksheets enforced by drill sergeant math teachers.
While I'm ranting, I checked out a book from the library yesterday called "Math with Bad Drawings", it's very fun, and approachable for anyone with no math background, kids and adults enjoy it.
We need more STEM for fun, and not just STEM for money. That's how we get good at STEM.
TFA says "How Anthropic uses AI to write 90-95% of code for some products and the surprising new bottlenecks this creates".
for some products.
If it were 95% of anything useful, Anthropic would not still have >1000 employees, and the rest of the economy would be collapsing, and governments would be taking some kind of action.
> If it were 95% of anything useful, Anthropic would not still have >1000 employees
I think firing people does not come as a logical conclusion of 95% of code being written by Claude Code. There is a big difference between AI autonomously writing code and developers just finding it easier to prompt changes rather than typing them manually.
In one case, you have an automated software engineer, and may be able to reduce your headcount. In the other, developers may just be slightly more productive or even just enjoy writing code using AI more, but the coding is still very much driven by the developers themselves. I think right now Claude Code shows signs of (1) for simple cases, but mostly falls into the (2) bucket.
I don't doubt it, especially when you have an organization that is focused on building the most effective tooling possible. I'd imagine that they use AI even when it isn't the most optimal, because they are trying to build experiences that will allow everyone else to do the same.
So let's take it on face value and say 95% is written by AI. When you free one bottleneck you expose the next. You still need developers to review it to make sure it's doing the right thing. You still need developers to be able to translate the business context into instructions that make the right product. You have to engage with the product. You need to architect the system - the context windows mean that the tasks can't just be handed off to AI.
So, The role of the programmer changes - you still need technical competence, but to serve the judgement calls of "what is right for the product?" Perhaps there's a world where developers and product management merges, but I think we will still need the people.
Been using claude code almost daily for over a month. It is the smartest junior developer I've ever seen; it can spew high-quality advanced code and with the same confidence, spew utter garbage or over-engineered crap; it can confidently tell you a task is done and passing tests, with glaring bugs in it; it can happily introduce security bugs if it's a shurtcut to finish something. And sometimes, will just tell you "not gonna do it, it takes too much time, so here's a todo comment". In short, it requires constant supervision and careful code review - you still need experienced developers for this.
Not to mention almost all the breakthroughs on top of the initial Web happened in the US: Mosiac, Netscape, Apache, Yahoo, Google, etc. Many of them started out as "non-essential" research projects.
That's the same company twice, Apache which was just another standard implementation of httpd (the cern thing) and one company that isn't and wasnt relevant outside of the us and surely isn't known as tech driver around here (Yahoo)
Actually a great summary of why the world does not actually blindly thinks USA! When they think about tech advancements
Apache was quickly spun off from its NCSA roots to become its own thing. There's a lot of history here that you're twisting around or ignoring. I'm not even American but, while lots of other countries made their own contributions, it's insane to argue that networked computing would have advanced anything like it did without the US's innovations borne of their (former) attitude of experimentation and exploration.
Edit: Also, the more competent models (Opus/ Sonnet to a lesser degree) are good at very complex subtask delegation that it can blow through and attempt and then verify in seconds, so not sure hand crafted regex examples are the best counter examples here
New code patch models that I didn't even take seriously are actually really impressive and pretty new
Parachute Sam into an island of cannibals, come back in 5 years, and he'll be king. Unless, of course, one of the cannibals is Mark Zuckerberg; then he might just get eaten.
OpenAI had a lot of goodwill and the leadership set fire to it in exchange for money. That's how we got to this state of affairs.