Akhil Deo

AI policy @ | Tweets about the geopolitics of technology | Prev: @flixrisk, @orfonline

[Pranay Kotasthane @pranaykotas]

How will the US AI licensing regime (subtly called AI diffusion) impact India?

Here's Take 1 #thread

[Surabhi Agarwal @SurabhiA_ET]

My break in today's ET:

India has found success in winning over major countries to make Global Partnership on Artificial Intelligence (GPAI) the apex body for all matters AI; tough negotiations going on, final announcement likely today evening!

https://t.co/uygxTKbEu0

[Akhil Deo @akhildeo94]

Hoping to see more reporting on China’s proposition for model evaluations and testing 👇🏽

[Yi ZENG @yi_zeng]

Introducing the Chinese AI Safety Network. https://t.co/AHc2m08K1x A complementary approach compared to the national AI Safety Institute in other countries. It is a cooperation platform for AI Safety across China that brings together various efforts related to AI Safety and Security at all levels, and that serves as a platform for dialogue, mapping, interoperability, and collaborations, within China, and that connects and contributes to the world. Current efforts from China that have joined the Chinese AI Safety Network include Chinese Academy of Sciences (CAS),  China Academy of Information and Communications Technology (CAICT), Peking University (PKU), Tsinghua University (THU), Shanghai AI Lab, Beijing Academy of AI (BAAI), China Information Technology Security Evaluation Center, and Center for Long-term AI (CLAI), etc.

[Ben Nimmo @benimmo]

Out today: @OpenAI's first threat report into covert influence operations using generative AI.

Some familiar ops in the mix, incl. Doppelganger and Spamouflage.

But no sign of meaningfully increased engagement from AI use.
https://t.co/EUL1vHY3lt

[Akhil Deo @akhildeo94]

The US DoC is reportedly in the early stages of formulating export controls on AI models! https://t.co/JFReGtKLeO

[Akhil Deo @akhildeo94]

👇A very interesting comparison of the acceptable use policies of different foundation model providers:
https://t.co/iPxTQWN9GO

[Caiwei Chen @CaiweiC]

Ahead of a potential TikTok ban, we spoke to over a dozen current employees to reveal what it's really like working there. Yes, it's more Chinese than you think, and it's less about politics and more about profit. My latest with @violazhouyi:
https://t.co/VJzqhMCHHs

[Akhil Deo @akhildeo94]

@j2bryson @Miles_Brundage Yes - I'm looking forward to it!

[Kumar Sambhav @Kum_Sambhav]

Following our investigation on how faulty govt algorithms denied food to hundreds of thousands of poor and declared living people as dead in India, @AmnestyTech today has released a technical report on the audit of one of the algorithmic systems (1). https://t.co/DErkurCj2o

[Suraksha P @SurakshaPinnuET]

IndiaAI Compute Capacity, envisions to erect AI computing infra by deploying 10K GPUs via PPP.

"The govt can do a pooled procurement negotiation on behalf of the Indian ecosystem & let the private players buy at that rate": @tanujb

https://t.co/KI8iwe0A3d
@EconomicTimes

[Akhil Deo @akhildeo94]

Excellent overview of China's response to US semiconductor export controls: https://t.co/JTpBilSOmU

[Akhil Deo @akhildeo94]

👇Super interesting piece! Are model weights constitutionally protected speech?

[Lawfare @lawfare]

"[M]odel weights are not generally used to express ideas among individuals but, rather, serve primarily as instructions to order machines to act."
@ARozenshtein pushes back on the argument that model weights for AI are protected by the First Amendment. https://t.co/V0lkcxZGy5

[Soumyarendra Barik @imsoumyarendra]

🚨You get a call from an MP candidate, addressing you by name, seeking your vote - except, they never made the call, or said the name. Parties are exploring AI for such messaging ahead of LS polls, revive dead leaders

I have a special story @IndianExpress https://t.co/GfIzODydRX

[Ruihan Huang @ruihan_huang]

🚨BREAKING🚨 AI cannot be built without talent. In an increasingly competitive AI landscape, how is talent flowing, especially in the post-pandemic era?

By examining 1,500+ career profiles, @AJCortese_ and I updated our Global AI Talent Tracker 2.0 and found:
1️⃣The US remains the top destination for top-tier AI talent to study and work;
2️⃣China has seen an increase in its domestic AI talent pool;
3️⃣The capacity of countries such as India to retain talent is expanding, with AI talents showing less mobility than 2020.

We have also answered many million-dollar questions, comparing across regions and over time. Check out our full report at @MacroPoloChina @PaulsonInst and track the elite talent behind the AI boom! https://t.co/WMYC1eTmbI

[Soumyarendra Barik @imsoumyarendra]

🚨Breaking/Scoop: The Cabinet today is expected to clear a Rs 10,300 crore AI Mission under which the government will offer viability gap funding to firms looking to set up AI compute capacity in India. More details @IndianExpress https://t.co/DadoU5aUh5

[Fabricated Knowledge @fabknowledge]

One of the better articles / graphics I've seen about making a chip in awhile.

https://t.co/R9u8rEG7BU

[Akhil Deo @akhildeo94]

🤦‍♂️

[Surabhi Agarwal @SurabhiA_ET]

Govt's missive that GenAI must seek permission before deployment has sent SHOCK WAVES among firms building LLMs, esp startups who feel it is "anti-innovation and not forward-looking".

Must read piece by @SurakshaPinnuET & @diarekhiET & @annapurnarET

https://t.co/8xy1Ho7CHN

[Arun Sukumar @arunmsukumar]

Three years ago, @akhildeo94 and I examined plausible pathways for China to interfere in India's electoral processes. We soon realized the market almost identically placed - re: social and political economy faultlines - for such interference is the US. https://t.co/I1Hvn0hxyw

[Akhil Deo @akhildeo94]

@ARGohdes @OUPAcademic Congratulations, Prof. Gohdes!

[Akhil Deo @akhildeo94]

National security and existential risk are increasingly overshadowing ethics and human rights as the driving force behind AI governance in Western capitals.

Some screenshots from the new White House Executive Order, UK AI Safety Summit and the G7 Hiroshima Process: