Siddhant Chatterjee

AI ethics and other spicy stuff

[Vir Das @thevirdas]

I flew Air India home from London. It was all great. The crew, happy to report, still the kindest in the sky. @airindia

[Neil Renic @NC_Renic]

I am once again pitching my romantic comedy:

- two academics start dating
- discover they are each other's terrible reviewer
- hijinks ensue

Working title: Love is Double-Blind

[Siddhant Chatterjee @sidschatter]

@Its_Badshah @raju_innocentev Absolute scum of the earth

[Rahul Desai @ReelReptile]

Necessary writing about the Rashmika Mandanna syndrome, and women un-telling women's stories, by @MadameSengupta:

https://t.co/OfVi1ZSacK

[Ethan Mollick @emollick]

As a rule of thumb, AI has not yet caused any major changes to labor markets that would be visible on charts. The only significant changes so far have been to freelance markets. That likely will change, but any pattern so far is due to other macroeconomic factors.

[techAU @techAU]

Ok Adobe's Firefly text to Video AI is better than OpenAI's Sora.

[Chris Do @theChrisDo]

First impressions of @Adobe #Firefly …
freaking amazing, all in one solution, geared for creatives, built using images they own!
What can it do? What can't it do is a better question.
Text to image
merge images
upscale
swap out images
It's good with (drumroll)… type! ← yas!

[Heather Cooper @HBCoop_]

Adobe Firefly Video (beta) is available for everyone now, and it is good!

Image to video:

[Siddhant Chatterjee @sidschatter]

@Klonick Can’t wait to read your op-eds on this, Kate!

[Anka Reuel | @ankareuel.bsky.social @AnkaReuel]

🚨 NeurIPS 2024 Spotlight
Did you know we lack standards for AI benchmarks, despite their role in tracking progress, comparing models, and shaping policy? 🤯 Enter BetterBench–our framework with 46 criteria to assess benchmark quality: https://t.co/8WJOVLPHnB 1/x

[Alexandru Costin @acostin]

Welcome to the world, @Adobe Firefly Video model (announced today, public beta later this year)! Designed to be safe for commercial use, for great cinematic quality and fluid motion, camera controls and of course, deep integration into our tools. The labor of love of a team dedicated to help creative professionals ideate and edit video. I can't wait to see what the community will do with it later this year! https://t.co/FvjeF8P5zQ #AdobeFirefly @creativecloud.

[Nik Marda @nrmarda]

Maybe the 100 most influential people in AI are actually 100 data labelers from the philippines

[Gary Marcus @GaryMarcus]

Unbelievable. @elonmusk took this entirely out of context and didn’t bother to do three seconds of research, and used his platform to imply that this was business as usual.

They were talking to blind people. FFS, @elonmusk.

[Arvind Narayanan @random_walker]

Here's my thesis on what went wrong with generative AI from a business perspective and how that's changing now.

When ChatGPT launched, people found a thousand unexpected uses for it. This got AI developers overexcited. They completely misunderstood the market, underestimating the huge gap between proofs of concept and reliable products.

This misunderstanding led to two opposing but equally flawed approaches to commercializing LLMs.

OpenAI and Anthropic focused on building models and not worrying about products. For example, it took 6 months for OpenAI to bother to release a ChatGPT iOS app and 8 months for an Android app!

Google and Microsoft shoved AI into everything in a panicked race, without thinking about which products would actually benefit from AI and how they should be integrated.

Both groups of companies forgot the “make something people want” mantra. The generality of LLMs allowed developers to fool themselves into thinking that they were exempt from the need to find a product-market fit, as if prompting is a replacement for carefully designed products or features.

OpenAI and Anthropic’s DIY approach meant that early adopters of LLMs disproportionately tended to be bad actors, since they are more invested in figuring out how to adapt new technologies for their purposes, whereas everyday users want easy-to-use products. This has contributed to a poor public perception of the technology. (A point we make in the AI Snake Oil book: https://t.co/foQpEhRN70)

Meanwhile the AI-in-your-face approach by Microsoft and Google has led to features that are occasionally useful and more often annoying. It also led to many self-owns due to inadequate testing like Microsoft's early Sydney chatbot and Google's Gemini image generator. This has also caused a backlash.

But things are changing. OpenAI and Anthropic seem to be transitioning from research labs focused on a speculative future to something resembling regular product companies. If you take all the human-interest elements out of the OpenAI boardroom drama, it was fundamentally about the company's shift from creating gods to building products.

Google and Microsoft are slower to learn, but my guess is that Apple will force them to change. Last year it was seen as a laggard on AI, but it seems obvious in retrospect that the slow and thoughtful approach that Apple showcased at WWDC is more likely to resonate with users.

Still, we shouldn't expect changes overnight. There are unsolved research challenges when it comes to making LLM-based AI assistants that actually work: https://t.co/T70F9QiheN

Enterprise adoption will likely be even slower. the barriers include integrating it into existing products and workflows and training people to use it productively while avoiding its pitfalls. We should expect this to happen on a timescale of a decade rather than a year.

If you want clear analyses of AI that look past the hype, subscribe to the AI Snake Oil newsletter. https://t.co/Esw1fkBrAm

[Yixin Wan @CVPR2026 @yixin_wan_]

🚨WARNING: Instructing T2I models to depict “diverse” people will severely harm historical factuality!🤯In this work, we benchmark the evaluation of this “FACTUALITY TAX” of diversity intervention prompts, and propose Factuality-Augmented Intervention (FAI) to resolve the issue.

[Leonie Koessler @Leo_Koe_]

🎉 Just uploaded a new paper “Risk thresholds for frontier AI” that I wrote together with my colleagues @jonasschuett and @Manderljung from @GovAI_.

Frontier AI systems could pose increasing risks to public safety and security. But what level of risk is acceptable?

More in 🧵

[Siddhant Chatterjee @sidschatter]

@CatWorkers @Ana_Alania

[Sebastian Farquhar @seb_far]

Is your LLM hallucinating? 👻

Our @Nature paper shows how to detect when an LLM is making things up.

A 'confabulating' LLM answers with inconsistent meanings when re-asked the same question. We use this to estimate uncertainty and detect confabulations.

Learn more 🧵👇 1/

[Laura Weidinger @weidingerlaura]

New paper out! Very excited that we’re able to share STAR: SocioTechnical Approach to Red Teaming Language Models. We've made some methodological advancements focusing on human red teaming for ethical and social harms. 🧵Check out https://t.co/TOTqh6HDPH

[Apar Gupta @apar1984]

Ranveer, the problem isn't your political leanings but an amoral ignorance, which you claim as an excuse, if not a virtue, for your actions. You are neither left, right, nor center. The truth is your leanings if any, are momentary, and wholly motivated by self-interest.

In the digital ecology, it means obeying the 'law of reach.' Here, views, average watch time, and celebrity co-branding matter more to you than framing a conversation that advances social welfare. The issue with such an approach is that it avoids all norms of media ethics and taking ownership on the platform you provide to your guests.

What else explains -- a video titled, "Career Hack Used by Hitler"? An association/collaboration with MyGov for interviews with Cabinet Ministers without a clear notice to your viewers? I hope you engage with criticisms such as mine, even response videos from folks such as @Memeghnad. Your audience, scores of whom are young Indian men deserve an improved standard of discourse irrespective of whatever political leanings you hold. Own up and take accountability for your work.