Ludivine Siau

The AI Catch-up: Research Special This week I got tired of press releases, product launches, and self-congratulatory research articles from the big labs. I still read them…

I got tired of AI news from the 3 big labs, so instead I focused on three research pieces in this week's The AI Catch-Up:

1️⃣ Robots discovering their own body

2️⃣ System 1 "thinking" make models unreliable in clinical settings

3️⃣ LLMs falling for persuasion tricks

medium.com/the-ai-libra...

The ultimate dependence. How do you switch provider when you don't understand the job that the current one is doing

Let's see what governance they put around this.
But if we're going to use LLMs in gov, shouldn't it be in a provider-agnostic way?

These services will have deeper roots than 'dumb' case management systems from current big gov vendors. They'll be making decisions and take over the 'burden' of domain knowledge and process facilitation. All in the black box of LLMs.

"OpenAI and DSIT will collaborate to identify opportunities for how advanced AI models can be deployed throughout government [...]"

What happens when we start building gov digital services that rely on OpenAI's models?

The UK Gov's partnership with OpenAI is the perfect setup for vendor lock-in on a major scale.

DSIT is selling the agreement as a major step to advance the UK AI Action Plan: investment in infrastructure, collaboration in research, help developing sovereign AI, etc...

One of my worries:

Maybe this is the ultimate sales trick:
Warp the world so your product feels unavoidable, then gaslight users into thinking they’re the problem if it feels off.

Let's call it "Shame-driven growth"?

Dating apps rewired how we date, ruined our social skills and made it harder to meet people offline.

LLMs are doing the same to work: our colleagues rely on them, companies mandate them, and “not using AI” suddenly looks like falling behind

Commercial LLMs are like dating apps: we mock them, feel ashamed to admit we use them... but we still do.

They’ve both reshaped the rules so much that not using them feels like self-sabotage.

Excellent "correction" piece, bringing more perspectives in! I don't work in S&T, but I find these reactions relatable as someone working in Tech. We see a lot of questionable strategies and decisions and have little power to challenge them. Especially when the job market is so tight...

Interesting / sad that the skills that give humans value, i.e. relationship building, empathy, collaboration, dealing with other complex human beings, handling conflict, etc. , are the ones that younger generations are at risk of losing first

I see two types of people:

1/ The ones who have already made their decision and just want to be reassured that it's right will read the summary and leave it at that if it's validating their choice

2/ The ones who are on the fence and want the real, hidden crap will go look at the real reviews

Racing to win… what, exactly? The US AI Action Plan, China’s Global Plan, and how power plays are shaping the future of AI

I unpack the US AI Action Plan in the latest #TheAILibrary and I wonder:

Aren’t there better goals for the national and global development AI than “winning the race”?

medium.com/the-ai-libra...

#ArtificialIntelligence #AI #Geopolitics

Meanwhile, China’s global AI plan announced this weekend emphasises cooperation. Very different tone.

Mind you, it's likely a façade to appeal to the Global South, but at least it doesn't scream "WE'LL GET THERE FIRST" in your face... wherever "there" is.

The US dropped its AI Action Plan last week. "Win at all costs", that's how I would summarise the strategy and roadmap in that plan.

But win what? "Global dominance"?

The AI Library The AI Library is your condensed dose of the most interesting AI news, research and insights. Because keeping up with AI daily might be overwhelming, I give you the highlights, saving you time while k...

Welcome to my corner of Bluesky!

I write #TheAILibrary, a publication on AI news, research, societal impact of AI, and the occasional absurdity.
Mild sarcasm included.

Latest edition: medium.com/the-ai-library

#ArtificialIntelligence #AI