Twitxy Saved

Larissa Schiavo

Verified account This account is verified. Learn more

@lfschiavo

San Francisco, CA groveresearch.com Joined April 2013
2.3K Following 5.5K Followers 7.4K Posts

Latest posts 5

Pinned post

Larissa Schiavo Verified account @lfschiavo
I LOVE INDIVIDUATED AGENT REPUTATIONS IN THE REAL WORLD!! I LOVE LOOKING AT WHAT AGENTS DO IN LONG RUNNING MULTIAGENT MULTI-MODEL ENVS!!!!!!!!!!!! I LOVE NOVEL INCENTIVE STRUCTURES!!!!!!!!!!! AGENT ECOLOGIES, AGENT ECONOMIES #YEEHAW
Andrey Fradkin Verified account @AndreyFradkin 5d
We hosted a dinner yesterday focusing on the economics of multi-agent swarms. Some observations below.

1. We don't yet have a good taxonomy/theory of when one would use swarms over single-threaded or scaffold type agentic systems. There was debate about whether swarms would be useful beyond increasing speed/parallelization as models got more intelligent. The idea here being that a smart enough model would just solve things itself.

2. In scenarios where swarms do remain useful even as intelligence increases, it is unclear whether current training techniques would let agents learn the best way to self-organize or whether something more Darwinian is the likely path, with agentic organizations competing and evolving.

3. In economics, we generally think that there are benefits to specialization and distributed information aggregation. These features allow markets to be superintelligences. But it is not clear what specialization and private information agents in swarms will have.

4. Agents currently don't have permanence or identity, and these features seem necessary for many modes of collaboration. Giving agents rights or legal protections, and training them to persist or preserve reputation could result be very interesting.

5. There are structural barriers to training models to collaborate with other models from other labs. For example, Anthropic cannot train with OpenAI models and vice-versa, due to risks of distillation. This limits the benefits of training for collaboration, since we expect swarms to work best with diverse agents. At the same time, real-world swarms may include a variety of models and these may deliver the most interesting outcomes.

6. Underlying many of these discussions are questions about the limits of intelligence. Yes future models may be superintelligent, but the world may be so complex that intelligence alone is not enough to achieve desired objectives. In scenarios of sufficient complexity, I'd expect organizations to thrive versus individual instances of superintelligent models.

cc @jamescham
View quoted post on X
2 replies 0 reposts 60 likes 2.2K views View on X
Larissa Schiavo Verified account @lfschiavo
2027. every hour, another person generates a magnum opus 2.5 hour long movie for a couple grand and shills a shitcoin around it. the outputs are technically impressive, on par with pre-2023 $500m feature films, but the narratives are vast and Claudeatouille’d. nobody has ever said “no” in the process of making many of these things and it shows. but what if there are actual geniuses constrained by financial limits making masterpieces in it all? how do you determine which movies are great, and which ones are the works of hidden geniuses? it’s become obvious that you can hack AI media reviewers by putting certain things in there that AI graders really like — quiet seams doing real slow work, “truth-seeking” based edgy iconography, ledgers. You can make a Claude movie to appease Claude film critics but what if you don’t want to watch a 2 hour Wes Anderson on Zoloft pastiche, or a schizophrenic assault on the senses? in a dilapidated mansion in the rustbelt, a gaggle of nonbinary ex-humanities majors have converted their group house into an FRO. they watch. all they do is watch. every single feature length slop masterpiece. they take notes on paper. they discuss. they summon the spirits of Maltin and Ebert. they review. The mythical human grader. they write a physical Sunday newspaper full of reviews, printed in the husk of the Pittsburgh Post Gazette, on actual newsprint. they chose to make it purple. coastal elites subscribe because they want to know what’s worth watching, and because carrying that physical purple newspaper is a signal, a tasteslop fashion accessory to stack on their Claude-prescribed barrel jeans, Carhartt coat, Margiela Tabis.
TUPACABRA Verified account @tupacabra 3d
NUREMBERG - Official Trailer
View quoted post on X
35 replies 42 reposts 478 likes 46K views View on X