Alberto Maria Pepe

RS @Meta (FAIR) | ex @Microsoft, @Huawei, @Cambridge_Uni

[Alberto Maria Pepe @albertomariape]

Happy to share that I have joined @Meta in their London King's Cross office, working on AI Research Agents within the FAIR team! 🚀

[Babak Rahmani @babakRmni]

1/ 🔍
For most of scientific history, we studied hidden mechanisms through behavior.
We could not directly inspect the system. We observed what it did, formed hypotheses, designed interventions, and inferred what was happening underneath.
Can AI agents do the same?

[Alberto Maria Pepe @albertomariape]

@matteo_achilli magari parametri...e trillion non si traduce con trilioni ma mille miliardi...

[Alberto Maria Pepe @albertomariape]

@matteo_achilli uso improprio? Non so con che coraggio avete rilasciato una cosa del genere

[elvis @omarsar0]

NEW paper from Meta.

(bookmark it)

It's an agent system that autonomously discovers neural architectures that beat Llama 3.2 at 350M, 1B, and 3B scales, all under a 24-hour compute budget.

They get this work by splitting the search into two agents:

> AIRA-Compose searches the macro architecture.

> AIRA-Design implements the low-level mechanisms.

For devs:

If one agent in your stack is doing both strategy and implementation, split it. Run a planner that picks the structure and an implementer that fills in the mechanisms.

AIRA shows this beats a single end-to-end agent on a real, non-toy search problem. The same split is useful for pipeline assembly, query planning, prompt scaffolding, and tool-use programs.

Paper: https://t.co/CYALI6CFjJ

Learn to build effective AI agents in our academy: https://t.co/1e8RZKs4uX

[Despoina Magka @MarlaMagka]

(🧵)Excited to share our latest work on AI Research Agents discovering novel language modelling architectures that show competitive performance when scaled up at 1B parameter size: https://t.co/IbW4LwMwu4

🤖 We gauge the ability of AI systems to autonomously design foundation models beyond the standard Transformer paradigm, by empowering LLM agents to perform both
🔍 high-level architecture search → AIRA-Compose
🛠️low-level mechanistic implementation →AIRA-Design

[Alberto Maria Pepe @albertomariape]

With 𝐀𝐈𝐑𝐀-𝐂𝐨𝐦𝐩𝐨𝐬𝐞 and 𝐀𝐈𝐑𝐀-𝐃𝐞𝐬𝐢𝐠𝐧 we show that agents are fully equipped to ideate their own successors. Read it here: https://t.co/0qZpNRSHXw

[Alberto Maria Pepe @albertomariape]

🤖📷𝐂𝐚𝐧 𝐀𝐈 𝐫𝐞𝐬𝐞𝐚𝐫𝐜𝐡 𝐚𝐠𝐞𝐧𝐭𝐬 𝐝𝐢𝐬𝐜𝐨𝐯𝐞𝐫 𝐭𝐡𝐞 𝐧𝐞𝐱𝐭 𝐠𝐞𝐧𝐞𝐫𝐚𝐭𝐢𝐨𝐧 𝐨𝐟 𝐟𝐨𝐮𝐧𝐝𝐚𝐭𝐢𝐨𝐧 𝐦𝐨𝐝𝐞𝐥𝐬? We put them to the test with two complementary, model-agnostic frameworks: 𝐀𝐈𝐑𝐀-𝐂𝐨𝐦𝐩𝐨𝐬𝐞 and 𝐀𝐈𝐑𝐀-𝐃𝐞𝐬𝐢𝐠𝐧 -- a thread:

[Alberto Maria Pepe @albertomariape]

Our agents designed custom attention mechanisms to handle long contexts performing within 3% of the human state-of-the-art, and optimized a nanochat training script under a 5-minute time budget, surpassing the original Autoresearch minimum validation metric.

[annie @_annieversary]

what the Fuck https://t.co/jSWRRiEBQ5

[Alberto Maria Pepe @albertomariape]

@reportrai3 immagina dover andar a pescare una foto ad un evento pubblico di 7 anni fa...

[Alberto Maria Pepe @albertomariape]

@CinghAmanuense @ultimoranet semplicemente no, non funziona

[Alberto Maria Pepe @albertomariape]

@lgf47264 @ultimoranet @cavicchioli forse hai perso un pezzo -- l'iran è una teocrazia che dichiara apertamente morte all'america e ad israele, sponsorizza il terrorismo internazionale e manda a morte i propri cittadini, l'ucraina no

[Wes Bos @wesbos]

Claude Code leaked their source map, effectively giving you a look into the codebase.

I immediately went for the one thing that mattered: spinner verbs

There are 187

[Stefan Schubert @StefanFSchubert]

While social media is polarising, evidence suggests AI may nudge people towards the centre.

This holds true of all studied models. Grok is more right-leaning than other models, but also has depolarising effects.

By @jburnmurdoch.

[Alberto Maria Pepe @albertomariape]

@DarkCigars imbarazzante

[Alberto Maria Pepe @albertomariape]

@GiuseppeConteIT ti sei ridotto ad un populismo veramente becero, fai pena e quasi un po' di compassione