Hacker News new | past | comments | ask | show | jobs | submit | hi's comments login

> 8.2 Natural Sciences Red Teaming Assessment Summary

"Model has significantly better capabilities than existing models at proposing and explaining biological laboratory protocols that are plausible, thorough, and comprehensive enough for novices."

"Inconsistent refusal of requests for dual use tasks such as creating a human-infectious virus that has an oncogene (a gene which increases risk of cancer)."

https://cdn.openai.com/o1-system-card.pdf


BUG: https://openai.com/index/reasoning-in-gpt/

> o1 models are currently in beta - The o1 models are currently in beta with limited features. Access is limited to developers in tier 5 (check your usage tier here), with low rate limits (20 RPM). We are working on adding more features, increasing rate limits, and expanding access to more developers in the coming weeks!

https://platform.openai.com/docs/guides/reasoning/reasoning


I'm in Tier 4, and not far off from Tier 5. The docs aren't quite transparent enough to show that if I buy credits if I'll be bumped up to Tier 5, or if I actually have to use enough credits to get into Tier 5.

Edit, w/ real time follow up:

Prior to buying the credits, I saw O1-preview in the Tier 5 model list as a Tier 4 user. I bought credits to bump to Tier 5—not much, I'd have gotten there before the end of the year. The OpenAI website now shows I'm in Tier 5, but O1-preview is not in the Tier 5 model list for me anymore. So sneaky of them!



> "o1 models are currently in beta - The o1 models are currently in beta with limited features. Access is limited to developers in tier 5 (check your usage tier here), with low rate limits (20 RPM). We are working on adding more features, increasing rate limits, and expanding access to more developers in the coming weeks!"

https://platform.openai.com/docs/guides/rate-limits/usage-ti...


I have tier 5, but I'm not seeing that model. Also API call gives an error that it doesn't exist or I do not have access.


I'm talking about web interface, not API. Should be available now, since they said "immediate release".



It may take a bit to appear in your account (and by a bit I mean I had to fiddle around a while, try logging out/in, etc for a bit) but it appears for me and many others as normal Plus users in the web.


Same for me here


The codebase heavily uses PyTorch XLA libraries (torch_xla.*), which are specific to TPU. Key TPU-specific elements include XLA device initialization, SPMD execution mode, TPU-specific data loading, and mesh-based model partitioning.

[0] https://github.com/felafax/felafax/blob/main/llama3_pytorch_...

[1] https://pytorch.org/xla/master/


hi @dang and team, why is my content flagged? It's legit and confirmed from multiple sources that it was a bad gateway.

Guidelines state:

"Please don't complain that a submission is inappropriate. If a story is spam or off-topic, flag it. Don't feed egregious comments by replying; flag them instead. If you flag, please don't also comment that you did."

What if it wasn't spam or inappropriate or off-topic then why flagged?

This confirms it: https://news.ycombinator.com/item?id=41257443

and this:

https://x.com/search?q=chatgpt&src=typed_query

Shouldn't the flagger of legit content get [flagged]? lol

Thanks team!


Tinygrad offers a superior MNIST implementation with minimal dependencies[0].

[0] https://github.com/tinygrad/tinygrad/blob/master/docs/mnist....


Autograd hides the backward pass in that implementation, but this code spells it out explicitly.


Anyone know a timeline for AMD on MLPerf?


It won't be for a while. It really takes someone to focus on this and it isn't just AMD. The team at MLPerf will need to step in as well and from my discussions with them, they are busy enough as it is with their own goals.

My company, Hot Aisle, has a box of mi300x (soon to be +16 more) that we have dedicated as a free resource to unbiased benchmarking. That's instigated articles like the Chips & Cheese one and the Nscale Elio post...

https://chipsandcheese.com/2024/06/25/testing-amds-giant-mi3...

https://www.nscale.com/blog/nscale-benchmarks-amd-mi300x-gpu...


AMD is already on MLPerf in the form of the tinybox red :)


They should have bought tiny for 600 million ;)


Now do #mi300x. I've already offered you the compute resources, but you called me an AMD shill, lol... pot kettle... ¯\_(ツ)_/¯


> "The persistent are attached to the goal. The obstinate are attached to their ideas about how to reach it."

I didn't know what the word "obstinate" meant so here you go: "stubbornly adhering to an opinion, purpose, or course in spite of reason, arguments, or persuasion."

While PG's quote suggests a clear distinction, it's overly simplistic. Persistence and obstinacy often overlap in practice, sharing traits like energy, imagination, resilience, good judgment, focus on a goal, and listening intently. The issue is that "reason" can be subjective. For example, Copernicus and Galileo were considered obstinate for his heliocentric theory, but history proved him right. This shows that the line between persistence and obstinacy is often drawn in hindsight.

Referencing the Collison brothers highlights a bias towards successful YC alumni. It would be more telling to classify current batch founders as obstinate or persistent and revisit their success in a decade.


> While PG's quote suggests a clear distinction

No it doesn't. The essay includes multiple parts talking about how the things are related, similar, sometimes indistinguishable, and also that it can be a spectrum.

In fact, arguably the entire thesis of the essay is how the two traits have both similarities and differences and that it is complicated.


> Persistence and obstinacy often overlap in practice, sharing traits like energy, imagination, resilience, good judgment, focus on a goal, and listening intently.

Obstinacy is defined by a lack of imagination, good judgement, and intent listening.

> For example, Copernicus and Galileo were considered obstinate for his heliocentric theory, but history proved him right. This shows that the line between persistence and obstinacy is often drawn in hindsight.

History didn't prove them right, science did. The fact that people considered them obstinate does not mean that they were. The only future where they would still be considered the obstinate ones is one run by obstinate people. They had the evidence, which was ignored by obstinate heliocentrists. Heliocentrists did not have convincing reasons for their belief that Copernicus/Galileo ignored.


> Obstinacy is defined by a lack of imagination, good judgement, and intent listening

I think that may be a mistake.

Any value strategy that is primarily conservative (e.g., protecting sunk or resource assets) will be obstinate. That doesn't make it slower or stupider.

So oil and timber companies and monopolists et al will keenly monitor opposition and respond immediately and deftly -- with reality-avoidance. As will individuals who are primarily guarding something they feel is at risk of being taken away.

They have the same or more intelligence, judgment, and active listening; it's just that their strategy is not creation or innovation.

Indeed, in a fair fight the innovator will lose to the conservative, because it's just plain harder to make things happen, particularly when it involves convincing others to change their patterns or minds.


The Consensys case exposes the fundamental mismatch between blockchain tech and securities law.

The SEC claims MetaMask Swaps 'effects transactions in securities,' but can a smart contract be a broker?

This isn't just about Consensys; it's about whether the '34 Act's definition of a broker can encompass code without breaking the entire DeFi ecosystem.

Congress's silence forces courts to decide if 'protecting investors' requires stifling the very innovation that could democratize finance.


Thanks @midzer! What was your high score?


I stopped the game when I had over 50 sols collected, I think. Preparing dinner now.

Gonna replay later again :)

What does the $ amount on the canvas tell me?


Oh nice! Try leaving it running while you make dinner :)

Bottom left $ is the value of each star system trust. Top Left $ is a high score = # sols collected x total trust value.

Backstory: the story goes that centuries ago a benefactor setup a $0.99 star system trust that compounds annually over centuries ($0.99 x 0.7)^648. So by the time a Solhopper crew arrives to collect the honeypot, each star trust worth ~$31,415,926,535,897,932.


Ok, I made to 100 in a fresh game with 605 sols passed. Need a break now :)

Really fun. Gonna share it!


Consider applying for YC's Spring batch! Applications are open till Feb 11.

Guidelines | FAQ | Lists | API | Security | Legal | Apply to YC | Contact

Search: