> The tech group has issued a force majeure notice related to the power supply to the project developer, a unit of Blue Owl Capital, after the site ran into permitting delays after rising local opposition.
Since when can failing to get a permit be considered force majeure?
Doesn't majeure specifically denote something large scale and out of anyone's control, such as a hurricane?
Oracle is in a pretty bad situation, it’s a desperate move. Their debt is rated just above junk since July. They are completely over leveraged. There is a reason the share price has been trending down so much
Its funny how before Oracle got in everybody was guessing who'd be swimming most naked when the tide goes out. CoreWeave obviously, but OpenAI likely as well depending on whom you'd ask.
With the contracts that "One Rich Asshole Called Larry Ellison" jumped on its without a doubt Oracle going down second if not first.
He's just gambling because he's old and feeling bold.
There's certainly been some small attacks in the EU+ caused by Russia, and many more attributed to Russia, and of course Ukraine, which is not in the EU and never has been, is bombing Russian cities in retaliation for the massive invasion and continued occupation Russia launched some years ago.
What makes to possible collapse of Oracle worse is that they have an actual business too that is currently getting cannibalized to feed the AI projects. If coreweave fails that is bad but oracle failing will be worse.
I hate Oracle and many of their products and long term it may be a blessing but a lot of conpanies, government agencies, hospitals, etc rely on their diverse suite of enterprise software.
Failing to get a permit is not an act of god, having to pay even when the site does t have power is poor milestone / payment planning. Also it’s quite hard to request payment if the thing you are buying is not working (appreciate this quite complex here), but if the people who are delivering the data center are responsible for delivering the power, it feels like oracle would be within reason to not pay?
Ellison has more or less lived his entire life running Oracle (and his personal finances) completely over-leveraged, skating just barely ahead of disaster. I don't see this ending any way other than SV picking apart its bones for pennies on the dollar.
We have to account for the fact that Oracle is a law firm with a software branch attached to it.
Investors who (rightfully) see this "force majeure" claim as bullshit will have to drag Oracle through court, and Oracle probably counts on this taking years to get settled.
Trumps blessing basically equivelent to stating AI is large scale fraud. Thats essentially what it means when he praises things: either he expects fraud or is congratulating.
You're wrong. The best-selling vehicle is the Flying Pigeon bicycle, the standard bicycle from Mao-era China with more than 500 million produced since 1950.
Depends on what counts in that 57 million. Just the base Corolla, or also the Corolla Verso that was actually based on the Avensis platform? What about the Toyota Auris? Toyota Conquest? Holden Nova? Geo Prizm?
I wonder if you combine the numbers of the VW Golf, Rabbit, Bora, Jetta, Corrado and Scirocco if it beats that 57 million.
The main question is whether to count the Corolla at all, given that it's changed so much over the years. The original VW Beetle at ~21M and minimally changed from 1938 to 2003 has a strong claim.
> It regularly seems to come up with terms and descriptions for things in its chain of reasoning and then uses these terms in its output assuming you understand what it's talking about.
GPT does this all the time, too (both Sol and Astra). I constantly have to tell it to not use terms that were not part of the initial prompt.
I don't know how people are managing hands-off with Sol/Astra over long horizon tasks. It will come up with 100 new concepts/terms and then use them as if you've been following every Edit call for 24 hours.
It’s probably an assumption baked into the training pipeline from that ancient time six months ago when we mostly were manually following each edit tool call.
> I constantly have to tell it to not use terms that were not part of the initial prompt.
Hah, yeah even when you put it in AGENTS.md or a skill.. constantly having to remind it.. "what does AGENTS.md" say about doing that?".. Thinking.. Thinking.. "Oh, it says I should never do that, I'll remember that next time.."
> In the post-AGI economy, humans will eventually stop writing
and reading code, but we still need an ambiguity-free way to
tell the AIs building the world around us what we want done.
> With laws, our intents can be much more precise than natural
language.
Doesn't this just mean that the code is now "laws", ie. the code is now the spec.
Given this, is there any reason think that writing the "laws" for a complex system is any easier than writing the old-fashioned code that implements it?
Yeah i think empirically you've hit the nail on the head (and there are some ties here to computability theory, e.g. Rice's thm).
sometimes the specification is easier to write than the code (sorting algo vs. quicksort impl) and sometimes the spec is much harder (what's "a good user experience"? what does "high availability" in a distributed system mean, precisely?)
i think it's just not true that it's easy to formally verify everything, it's often much easier to just write the code lol (e.g. sel4 is 200k+ lines of proof, ~50k lines of code iirc).
100% agree. It's crazy so many Ai articles are heralding the end of code while at the same time defining more complex ways to write code under a different name.
I suppose I could look it up but I wonder if people thought this of "high level" languages like C when it first came out. No more assembly. Or even assembly instead of machine code.
This is the age old problem with proving correctness. You can (sometimes) prove that two programs have identical behaviour. One of the programs can be slightly simpler in that it is only concerned with what the correct result of a given operation is, not how to get there. But this doesn't fundamentally change that the complete specification is almost as complicated as the program itself.
Exactly. You could simplify things by giving AI a limited set of laws, but then you'd risk that AI would make some mistake in the parts that you didn't cover.
So a failsafe set of laws would look very much like an actual program.
No, it doesn't assume that. It simply assumes that you can verify whether or not the LLM's implementation adheres to the laws you defined up-front — which it does not if it modified the laws.
Somewhat off topic, but I've started wondering if we can actually make local LLMs feel smarter than the frontier closed models, by post-training it for your specific use case.
Say Company X has a software product which consists of a million lines of code, including a ticket for every bug and new feature for this piece of software. Then wouldn't it make sense to try using an open weight model but post-train it on that specific code base while using the tickets to teach the model about past bugs and features. Not sure exactly how, but it could involve doing reinforcement learning solving a past bug on that historic version of the code base, and rewarding the model if it comes up with the correct solution (as defined by the linked PR which fixed the bug).
So essentially post-training your local open-weight model using reinforcement Learning with Verifiable Rewards (RLVR) on your software products history of bug reports and their ultimate solution. And the same for new features.
You could do this for sure but its quite a lot of effort and a normal software stack is quite well known due to the massive amount of github projects and other opensource code.
You can also get a lot out of a harness in this case or your Agents.md or Claude.md file by just enhancing the context.
You might even question yourself if you are doing something wrong if a modern LLM really struggles with your code.
We do the finetuning only on small semantic data were it helps a lot.
I'm still wondering when we see smaller models (faster and cheaper) for more specific stacks like spring boot + java + angular + english only or so. Interstingly enough, i assumed LLMs are really good in any language but it seems that non english languages do reduce the ooutput quality of an LLM. At least last years GTC there was a talk about it.
There are companies though which ahve this exact problem with programming languages you normally don't see. ABAP for example is a very well known SAP language.
So we have free text paragraphs which describe machines. Our model detects if the machine in question is of the main category we are looking for, then our finetuned model extracts from the free form text semantic information about the machine.
Like color, features, horsepower etc.
This would normally take quite a long time to do manually but we already had the semantic version of these texts because the company was doing this for a while.
We now use gemma or Qwen (we regularly re-finetune the newest models to just see if they get better and they actually do) and then use these finetuned models to save us a lot of time.
----
If I had a coding language which isn't available much online and coding LLMs are bad on it, I would definitly try to finetune this but its defintily a lot more work than just doing what I explained above.
Depending on what your usage of this APL Dialect is, it might be easier to fine tune it to migrate from your APL to something a lot more common.
If this is not an option at all:
You need to start creating data for the finetuning. You need a few hundred up to a few thousand of them in different formats like Q & A pairs. Documentation, syntax, a lot of diverse small examples.
The intersting thing about this data generation: You can either leverage, to a certain degree , what you already have, or you start collecting them through your team/work collegues or you do them by hand.
You can put the text into RAG and experiment with context engineering until a LLM is 'good enough' in it to be able to help you generating examples for and with you.
Like you give an LLM all the relevant context for it, then you let it generate pairs:
{
"instruction": "Write an expression to find the maximum along the rows of a 2D array.",
"input": "Array matrix: A",
"dialect_notes": "Custom dialect uses ⌆ (max-reduce) and ⌥ for axis specification instead of /[1].",
"output": "⌆ ⌥2 A"
}
(I have no clue about APL this is just a random example I asked an LLM to generate).
You might have luck and finding communities with the same issue you face.
I wonder the same thing. Code is pretty open ended though so I wonder if it’s not the best example.
On the one hand it would definitely be useful for something like classifying support requests into priority. But would it be worth it to just train your own model? I guess one advantage is you could give it some well informed guidelines without training something on lots of data.
There is probably a better example between discrete labeling and code though.
I'd like to have a model that just "knows" our code base and past issues, instead of having to query them for the same reason I don't want the model to have to query a dictionary to speak proper English.
I could be mistaken here, but I'm curious to see if it would improve both output speed and quality. Currently, every new prompt I start it basically spends 10-15 minutes "getting familiar with the code" which is both annoying to wait for and wasteful money-wise.
Since when can failing to get a permit be considered force majeure?
Doesn't majeure specifically denote something large scale and out of anyone's control, such as a hurricane?
reply