In era of pretraining, what mattered was internet text. You'd primarily want a large, diverse, high quality collection of internet documents to learn from.
In era of supervised finetuning, it was conversations. Contract workers are hired to create answers for questions, a bit like what you'd see on Stack Overflow / Quora, or etc., but geared towards LLM use cases.
Neither of the two above are going away (imo), but in this era of reinforcement learning, it is now environments. Unlike the above, they give the LLM an opportunity to actually interact - take actions, see outcomes, etc. This means you can hope to do a lot better than statistical expert imitation. And they can be used both for model training and evaluation. But just like before, the core problem now is needing a large, diverse, high quality set of environments, as exercises for the LLM to practice against.
In some ways, I'm reminded of OpenAI's very first project (gym), which was exactly a framework hoping to build a large collection of environments in the same schema, but this was way before LLMs. So the environments were simple academic control tasks of the time, like cartpole, ATARI, etc. The @PrimeIntellect environments hub (and the `verifiers` repo on GitHub) builds the modernized version specifically targeting LLMs, and it's a great effort/idea. I pitched that someone build something like it earlier this year:
https://t.co/ANHhasxzD8
Environments have the property that once the skeleton of the framework is in place, in principle the community / industry can parallelize across many different domains, which is exciting.
Final thought - personally and long-term, I am bullish on environments and agentic interactions but I am bearish on reinforcement learning specifically. I think that reward functions are super sus, and I think humans don't use RL to learn (maybe they do for some motor tasks etc, but not intellectual problem solving tasks). Humans use different learning paradigms that are significantly more powerful and sample efficient and that haven't been properly invented and scaled yet, though early sketches and ideas exist (as just one example, the idea of "system prompt learning", moving the update to tokens/contexts not weights and optionally distilling to weights as a separate process a bit like sleep does).
#OlympicGames - Inizia la progressione di Ginevra #Taddeucci. L'azzurra, autrice di un'ottima prova, incrementa il ritmo di bracciata, recuperando alcune posizioni. Adesso la toscana, in passato iscritta nel TNT Empoli, è situata in terza posizione.
Coulombic efficiency is an important metric when it comes to evaluating an EV battery’s performance and potential lifespan. Recently, we adopted new tools to improve the precision of our measurements, and last week at #AABC22, we shared new data on our Coulombic efficiency.
In our latest Solid-State University video, QS CTO @ironmantimholme delves into voltage stability in solid-state electrolytes and explains how we overcame the challenge of finding the right compounds for automotive-grade solid-state batteries. Learn more: https://t.co/9ujoW5mPiL
[DATI #RIVISTI] #terremoto ML 5.7 ore 07:07 IT del 09-11-2022, Costa Marchigiana Pesarese (Pesaro Urbino) Prof=8Km #INGV_33301831 https://t.co/oivMwtmPRd
Last week, for the first time ever, we presented a live demo of our extreme fast charging EV batteries on stage, in front of the huge audience at @WebSummit.
We are extremely excited about this milestone, and thankful for the amazing feedback from the crowd and the media.
@donush
Questa era la situazione odierna al lago di Lando a quota 1.400 metri con vista sulle Tre Cime di Lavaredo. Se non fosse per un timido accenno di foliage autunnale sembra un'immagine tipicamente estiva. Non c'è traccia di neve.
Oggi a Lomne', in Francia, si sono toccati i 33.3°C. Tutta Europa è alle prese con un'ondata di caldo anomalo che dura ormai da settimane ma finalmente s'intravede un cambiamento. Tra 4-5 giorni aria fredda dal Nord Europa porterà le temperature su valori più consoni al periodo.
Cell Architecture: We’ve developed a hybrid battery cell architecture combining what we believe are the key features of conventional pouch & prismatic cell formats. We believe these characteristics will enable our cells to overcome the limitations of existing formats.
Questo è il Monte Rosa a quota 4.500 metri. Solitamente in questo periodo dell'anno è già abbondantemente innevato dai 2.000 metri in su mentre quest'anno sotto i 4.000 metri non c'è traccia di neve.
We are excited to unveil the very first next-gen solid-state battery with 100% silicon oxide anode in the upcoming @mondialdelauto , the #ParisMotorShow!
Mark your calendar and join us in exploring the frontier of #solidstatebattery technologies!
"2022-10-07
NASA’s Solid-State Battery Research Exceeds Initial Goals, Draws Interest
NASA researchers are making progress with developing an innovative battery pack that is lighter, safer, and performs better than batteries commonly used in vehicles and large electronics today."