what machine are you running GPT-OSS-120B on? I'm currently only able to get GPT-OSS-20B working on my macbook using Ollama

I'm hoping their example is just the surface level and the technology is much richer. I'm very interested to see how they're adding: - real-time data to LLM retrieval - validation of recommendations (hallucination reduction) - enriched metadata for recommendations

Woah, this is wild. Never knew how aggressive the copyrights were for *Zilla.

When you consider the most recent Godzilla film, Minus One, is the 37th in that franchise, and was not only nominated for an Oscar, and may be it's most lucrative, you can see why Toho would aggressively police that copyright.

Minus One is legit a great film.

No, I can't. That makes it even less reasonable.

Edit: on the other hand, they apparently elected to use Godzilla as their logo, which kind of ruins the otherwise considerable unrelatedness.

Wonder how the people of Zillah Washington feel about it

I'm fairly sure that the antiquity of the Book of Genesis trumps any conceivable trademark claim!

It’s not called Boardzillah, so maybe not much at all.

Meant more about Godzilla copyright holders being litigious

Ah. I assume their trademark doesn’t extend to cities, but no doubt someone’s issued a legal threat and retracted it at least once!

Makes sense. And, TF Lite is excellent for on-device models and inference.

This seems bad. Heating of oceans will trigger a change in underwater currents, which will have a dramatic impact on heat distribution on the planet.

I've read somewhere that carbon emissions had declined for a few years. Is that true?

It's not true. Carbon emissions are record high and growing. I think I read that carbon emissions in the US declined somewhat this year. I also read (small sliver of hope) that we have avoided (so far) the worst-case climate scenario.

The US is also producing a record amount of oil (locked-in emissions), Biden just approved some huge drilling lease and they're considering allowing construction of new LNG terminals in the Gulf. :-\

Duolingo was one of the first companies to partner with OpenAI on leveraging their language models. I think they were using GPT-3 before ChatGPT was even popularized.

Generally agreed. I think 4-5 data points is fine, but the advantage of pie charts is that it's a quick visual representation of what proportion of the 100% can be attributed to the 4-5 segments.

Pie charts fail again once the distribution is very skewed (1 segment = 99% and the other 3 shared 1%). It also fails when the distribution is too even (5 segments 20% each). They shine the most when there is variance, but overall easy to interpret the distribution differences.

A new post-pretraining method for LLMs with an expansion of Transformer blocks.

I thought most vaccines were weakened viruses to prepare the body for diseases or viruses. None-the-less, very cool research.

Pretty sure this was a homework assignment I had while I was in school