This was, I believe, the problem that Microsoft wanted to resolve with their gradual burndown of WSUS - a lot of shops (including one I used to work at) would selectively roll out updates based on whether they thought they were relevant, resulting in an explosion of configurations that Microsoft had absolutely never tested against, and naturally, a lot of breakage.
It's not a sufficient criteria by itself, but where no better criteria is possible it would still produce better results in reinforcement learning than if the model has no reward for producing correctly compiling code vs code that failed to compile.
That's already happened. Its well established now that the internet is tainted. After essentially ChatGPT's public release, a non-insignificant amount of internet content is not written by humans.
Even electric signals are plenty fast enough. The issue as always is all the processing like modulation and demodulation. And ofc routing. Fibre does help with this as you can do longer distance without doing it.
Depends on the radius, as I understand it. It could have more or less gravity depending on that, but we don't know what the radius of this planet is yet.
Wolfram Alpha has a good calculator if you want to play around with this [0]
The main problem SDKs solve is providing a way to interact with an API without having to worry about nitty gritty details like serialization and deserialization, authentication, error handling, retries, connection and thread pooling, etc.
When you have a good SDK, interacting with an API feels like working with regular functions and objects in your chosen programming languages, so it _greatly_ simplifies integrating with an API.
Hey, OpenAI, so, you know that legal theory that is the entire basis of your argument that any of your products are legal? "Training AI on proprietary data is a use that doesn't require permission from the owner of the data"?
You might want to consider how it applies to this situation.
While I disagree with Ed Zitron's opinion that we are at 'peak AI', I do completely agree with his other points, especially the points he was making that OpenAI and Anthropic had no incentives to make things more efficient and therefore they were just drinking from the money hose. If anything it is shocking that nobody had done what DeepSeek did earlier. Good for them.
US unipolar hegemony has averted an all-out war, but it has also been very bloody, or at least immiserating, for people at its periphery.
Multipolarity doesn't imply we go back to the 1910s. The idea would be to strengthen multilateral institutions that put a check on things like the World Wars.
My going hypothesis is that the US cannot politically solve their Modern Monetary Theory national debt blackhole problem and is therefore itching for WW3 to be able to "reset" debts when the dust settles. Taiwan seems like a good opportunity for that, but maybe there is some sanity left in Congress.
I usually ignore people who say this because it guarantees they're not actually doing anything meaningful with these models and just want to squibble over semantics from the sideline...
They didn't just toss model weights over a fence, they shared exactly how to do what they did. They made a meaningful contribution that people are replicating with other models readily.
every time you respond to an AI model "no, you got that wrong, do it this way" you provide a very valuable piece of data to train on. With reasoning tokens there is just a lot more of that data to train on now
Earlier today I read a reddit comment[1] about a guy who tried running the quantized version from unsloth[2] on 4xH100 and the results was underwhelming (it ended up costing $137 per 1 million tokens).
I was a member of an STC chapter in grad school and I went to a bunch of meetings and networking events hoping to get some practical career advice from more experienced technical writers.
At no point did it feel like anyone in the org could (or would) actually help me in the transition from grad school to technical writing work. It was a really disheartening experience. It does not surprise me that membership has waned over the years.
In my experience, the biggest issue with US public schools is incompitent administrations. Those administrations ultimately report to school boards, whose members are selected via public election.
But how many people can name even a single person on their local school board? How many can account for what that person's platforms are regarding the school? How many can account for anything that person has done while on the board? Where would you even go to find out what your local school board is doing? How many of your school board members have observed the day-to-day goings on of your local schools?
The system seems fundamentally broken to me. It's no wonder school administrations are so incompitent - especially in larger districts.
Awesome - we'd love to have our CEO/CTO chat with you and your team if you're interested. Shoot me a note at mike.bilodeau @ baseten.co and I'll make it happen!
Thank you for all the work you guys do. The Arrow ecosystem is just absolutely incredible.
My few gripes related to interop with duckdb are related to Arrow scanning/pushdowns. And this extends to interop with other projects like pyiceberg too.
Registering an Arrow Dataset (or pyiceberg scan) as a "duckdb relation" (virtual view) is still a little problematic. Querying these "relations" does not always result in an optimal outcome.
For Arrow datasets, you can intercept the duckdb pushdown, but duckdb will have already "optimized" the plan to its liking, and any scanning restrictions that may have been more advantageous based on the nuances of the dataset might have been lost. Eg:
Perhaps in a similar way, turning an pyiceberg scan into a relation for duckdb effectively takes the entire scan and creates an Arrow Table rather than some kind of "scan plan" for duckdb to potentially make more efficient with its READ_PARQUET() functionality.
Most of this is probably dependent on duckdb development, but all of the incredible interop work done across communities/ecosystems so far gives me a lot of confidence that these will soon be matters of the past.
This reminds me of a (probably apocryphal) story about fast food chains that made the rounds decades ago: McDonald's invests tons of time into finding the best real estate for new stores; Burger King just opens stores near McDonalds!
