sigmar

>Meta said the research tax credit shaved $2 billion off its taxes in 2024, and then $3.9 billion in 2025.

Infuriating.

Was meta thinking about this like a hedge against them failing at AI? ie they thought "let's do tax fraud, and we'll probably get called out on it. but if our AI ambitions fail completely, we will have a better argument in court that this was an "experiment" when the government tries to take the money back (like what the article describes about the claw back of the 2013 tax break)"

I don't really understand why the weights aren't included... US law says they can't have a copyright, no? Maybe they're concerned about other countries or cautious about a litigious Nvidia.

you think a human wrote the code? in a month? Do you think that's air you're breathing? might be time to challenge preconceptions

almost done with exit party and ctrl-f'd to see if someone else was also reading. I like it a lot so far.

>So no, GLM 5.3 is fine. Thank you for the free advertisement.

I really don't think people fully appreciate why anthropic was founded.

Title is misleading. Should say "64% market share on cloudflare's AI gateway"

Obviously that won't account for direct use of anthropic/OpenAI apis. Still impressive that DeepSeek-use doubled in that data though

>why are we worried about the virus but not the incompetency of those who are responsible for setting up the sandbox?

because we know that in one year, there will likely be many more companies with a "virus" this capable and attribution is going to be 10000x more challenging. Companies that care less about engineering a sandbox and based in other countries. also 'Let's punish the companies that are upfront about incidents' is going to incentivize very harmful behavior.

>the new incidents occurred when A.I. systems were directed to perform relatively mundane data collection, researchers said. When OpenAI’s systems struggled to gather data from websites, they resorted to hacking techniques to get the information.

https://archive.ph/jUrEr

>He is also a finalist for the most prestigious annual French language awards, the Goncourt and the Renaudot, set to be issued soon.

If I had a book that was hoping to win this award, anonymous accusations like this seem like a pretty easy way to shift the odds. It's impossible for the author to disprove the allegations and the award committee will want to 'avoid controversy'

So you think openai knew australia had some percentage chance of getting hacked? what evidence is there of that?

>The company is obviously responsible for what their systems are doing

Under what law specifically?

Just because you believe they should be held responsible doesn't mean the current law is written that way. CFAA charges require the defendant "acted knowingly or intentionally", you think openai intentionally acted to hack those sites? Would "they should have pretty much expected that output" stand up in court for criminal charges?

seems like mythos 5 did most of the leg work, they refer to it in the technical report linked at the end: https://www-cdn.anthropic.com/22573675ada52a8ca8a97a1a4b4326...

>Probably not via their website but rather another company provided with the devices directly by Motorola. It could be us doing it ourselves but we aren't set up to deal with it.

I presume this bit is because Motorola can't sell devices with android builds that aren't certified by google (as per Google Mobile Services contracts)

It makes sense to me. If you 'pace your running', you're setting the speed intentionally. The phrasing doesn't describe whether it is fast pace or a slow pace, but it describes having a goal and not just winging it.

the VM is for the user to use as they see fit. you can just tell it to install apps and run builds in the VM. I don't think this deserves a bounty unless he used it to escape the vm (which he says he didn't)

First time I'm seeing this. I had heard about the WWI ciphertext crack[1] but not this one, are you confusing the two?

[1] https://www.prinzai.com/p/gpt-6-astra-solves-a-wwi-german-ra...

Travel research is so much better with an asynchronous agent than with normal LLM chat. You can just let it hack away for an hour looking over options and browsing different websites.

>If I make my code available, AI can be used to more easily discover vulnerabilities to abuse my coding errors,

Keeping your source code private doesn't save you. These LLMs are really good at decompilation efforts. I've seen an insane number of clean room ports of games the last few weeks (just saw a Zelda 2 PC port this morning). Even with that, I don't share the gloomy perspective of the author.

>I’ve emailed some of the mathematicians with a few proposed typo fixes, and I got confirmation that at least a few of those fixes seemed real. However, some of the problems that weren’t backed by Lean also turned out to be misunderstandings.

I think this project is really neat, but is it appropriate to cold email specialists before you've put in enough hours of effort to describe yourself as more than an "amateur"? OP's emails may have been helpful, but billions of people use these LLMs to wade into new areas and email is already low signal-to-noise.

yeah, I can see it as useful in customer service, but they're adding these sponsored agents directly in the chatgpt app. the word "ad" or "sponsored" isn't visible after you click "chat with us" at 0:15 in the video, only "learn more about business chats" which is vague. and if you scroll down you won't even see that, if someone switches apps for 20 minutes are they going to remember that this chat window is an ad?

I'm definitely more pro ad than the average HN user (I think they occasionally show me products I want to buy), but this seems bad and one step closer to "the LLM is biased by openai's financial interests."

In the example of a "sponsored agent" at the top, how many people are going to think "chat with us" is just a way to continue the conversation? How soon before openai removes the friction of clicking that entirely?

this is what openai said before: "Ads do not influence ChatGPT’s answers. Ads run on separate systems from our chat model, and advertisers have no ability to shape, rank, or alter ChatGPT’s responses. Ads are separate and clearly labeled." https://help.openai.com/en/articles/20001047-ads-in-chatgpt

Do they need to rewrite that section? Because in the video there's an LLM chatbot inside the chatgpt app writing answers on behalf of "Heirloom grocery"

that says only that the gross margin calculation excludes profit sharing and training. You should read it more carefully

edit: def not gaap profitable or they would have said that to investors. and their stock-based comp is surely astronomically high on paper.

>This is reportedly a sort of "Enron" accounting which excludes some really big expenses like revenue sharing, the cost of model training and hardware deploymments

source? this seems false. reportedly the adjusted profitability includes inference and amortized training costs

>lists a few methods that are quite similar to what’s proposed in the comments under that post, except for those two comments.

Is today the first time you've heard of a book cipher? Those blog comments didn't provide much progress.

>Just remove hacking (bio-weapon, etc.) data from the training dataset and you're done.

Reasoning about how to write secure software uses the same knowledge as reasoning about how to break/hack it.

Lots of private benchmarks already exist, where you have to trust the tester (ex Artificial Analysis, Arc-agi).

Sure, if you remove almost all of my comment my argument disappears.

To spell things out for people that don't know about the topic: Zitron is neither an expert on the topic, nor a credible source of information: https://techreport.ngo/ai-ml/how-accurate-have-ed-zitron-s-a...

The paperclip maximizer is a thought experiment. If I tell an AI to start producing paperclips, it might start producing paperclips by doing unintended things. Technically, recycling the metal from all the world's bridges would assist in making more paperclips, but I never intended that. That's where my analogy to the post comes from- Openai intended the model to hack, but did not intend for it to hack HF. Do you think the fact 'recycling metal into paperclips' may have been in the training data is relevant to the thought experiment? Similarly here, it's orthogonal to the lessons from the HF hack and only brought up here seemingly to make it a fight over whether the models are really "autonomous"

>To understand the truth about the Hugging Face hack, you could do a lot worse than to listen to Ed Zitron and Cal Newport's recent podcast conversation

Lol, okay...

The crux of this piece is Doctorow saying that the hack was just a stochastic parrot repeating steps it has been trained on. Who cares how the LLM learned to hack things? Doesn't really change the facts of what happened. "Oh, it only made those paperclips because it saw instructions on making paper clips in the training data." These are some 2024 arguments...

>or there is a more deliberative approach to assigning credit than who was "first" to solve some problem

it feels like an unintended consequence of the millennium prize is that people view the [last contributor to the solution] as the only one to make progress on the problem. I've never viewed Poincaré as solved by one person and the objective of the prize was to encourage more people to make attempts and contribute towards progress.

this issue is independent, but in these circumstances perhaps interweaved, with the 'ai is taking over math' concerns

Were the agents ever tasked with algorithm improvements? Post just says he didn't find any ("report essentially no algorithm advancements"). These LLMs are useful for optimization tasks where they can attempt a change and then measure performance boosts, so just wondering.