Logo
Picture

AP @AP798111

Don't cut your nose to spite your face Joined May 2020
  • Tweets
    9
  • Followers
    10
  • Following
    400
  • Likes
    252
AP retweeted
karpathy Profile Picture

Andrej Karpathy @karpathy

12 months ago

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: x.com/karpathy/statu… 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).

PrimeIntellect Profile Picture

Prime Intellect @PrimeIntellect

12 months ago

Introducing the Environments Hub RL environments are the key bottleneck to the next wave of AI progress, but big labs are locking them down We built a community platform for crowdsourcing open environments, so anyone can contribute to open-source AGI

AP retweeted
Snowden Profile Picture

Edward Snowden @Snowden

5 years ago

Nobody has stronger opinions about Joe Rogan than people who have never listened to Joe Rogan.

AP798111 Profile Picture

AP @AP798111

5 years ago

Thank you @beaucarnes for writing this helpful article. Learn React by Building an eCommerce Site freecodecamp.org/news/learn-rea…

AP retweeted
joshk Profile Picture

Josh Kopelman @joshk

6 years ago

Got this gem from @vkhosla: "If you prioritize for maximum risk reduction, you typically end up choosing a path where the consequences of success become inconsequential..."

AP retweeted
MetLife Profile Picture

MetLife @MetLife

6 years ago

Last night, the MetLife Building was shining bright in honor of All In NYC. � this tweet to show your love for NYC! #AllInNYC @NYCEDC @nycgo

AP retweeted
FGraillot Profile Picture

Florian Graillot @FGraillot

6 years ago

Great overview of #tech trends that have recently accelerated during the lockdown period, and what to expect next. #startup #InsurTech 👉 buff.ly/37xdOcS

AP retweeted
chrija Profile Picture

Christoph Janz 🕊 @chrija

6 years ago

RT if you (like me) start every spreadsheet with a silly narrow empty column and a silly empty row because your eyes hurt without some extra space.

AP retweeted
a16z Profile Picture

a16z @a16z

7 years ago

“Super apps will rescue us.� —@conniechan. But what is a super app? Dive into this a16z Summit talk to see why super apps are part of our future along with three other big trends in consumer tech. bit.ly/38wA4nd

Trends for United States

Good Saturday
Virginia State University
Happy Birthday Gary
Caturday
#DuangConcertDay1
Doom
Rogue
DUANG GO ROUND DAY1
#SaturdayVibes
Arday
DevOps
Introducing
X-Men
Cambridge
PERTHSANTA LOVE CLASS
#VIDALoveLessonxPerthSanta
#ashamed
#NationalRelaxationDay
Go Birds
Edrees Farooq

You might like

We power customer engagement with bespoke #AI technology that helps businesses to build valuable lasting relationships with customers.
EBO
@EBO_ai
278 Followers
ALpne Cet
@SumthinW
3 Followers
Dave $Omen🅾ï...
@DMccar3
1K Followers
Veteran Apple user. Fully immersed in the Apple world. I track my health with Apple, Withings and Omron devices. ⌚�🖥📱💻
Apple4ever🇫ð...
@Gryphon41737021
172 Followers
@wizkidayo die hard�� @chelseafc fans libra ♎�
Olanrewaju
@Olan_rewaju1610
155 Followers
Veritas Iustitiae
JaxonMohr
@JaxonMohr
90 Followers