Notices by Koen Hufkens, PhD (koen_hufkens@mastodon.social)

This is not about your copyright, safety or any human concerns.

This is a way to prevent the models from ingesting their own output, and limit scraping to only original human (high value) content.

Remember these people buy up old (human authored) books with the container load only to rip them apart to give their models a slight edge.

It is a way to steal your original content more efficiently.

https://futurism.com/artificial-intelligence/people-horrified-busted-anthropic-watermark

@aral @Gina @oscl @_elena @paolo hear hear, the reason I don't do academic conferences anymore - would be fun, can't afford it.

So, this "curing cancer" thing will have to wait to feed the word salad generator.

https://www.engadget.com/2225849/google-shuts-down-alphafold/

@cwebber Since for some (if not most) of these there should be absolutes, I don't see it improving things. I would say that the current foundation of the models doesn't allow for any of this. For reference, I read the magic wand as improvements on the current paradigm, not a new one.

Overheard two tourists at the local bakery:

"Now we know that we can't trust AI for the bus schedule".

#AI #search

@cdarwin https://mastodon.social/@koen_hufkens/116867334063441617

RE: https://flipboard.com/@theverge/the-verge-5hobbaa8z/-/a-RKgzzvvtTnadjMq82fXG_A%3Aa%3A43611565-%2F0

"intentional backdoor", read that again.

Don't buy "smart" appliances if you can, isolate if you must, shoot when it makes an unexpected noise.

The slow death of the power user.

"This isn’t an accident. This is the result of two decades of deliberate, calculated effort by the largest technology companies on earth to turn users into consumers, instruments into appliances, and technical literacy into a niche hobby for weirdos. They succeeded beyond their wildest expectations"

https://fireborn.mataroa.blog/blog/the-slow-death-of-the-power-user/

#technology #tech #sustainability

@EUCommission First of all, ban the Big Tech lobby. Then the rest will flow from there.

Arguably an improvement.

https://www.theverge.com/tech/910990/meta-ceo-mark-zuckerberg-ai-clone

We've lucked out tonight, with pressure from China. Now we have to see if the senile president can keep his temper in check for two weeks if met with resistance, and Israel doesn't kick the hornets nest.

@ai6yr @cwebber When mentoring students I often get the question - how do you figure things out so quickly.

Then I tell them that I've been messing with hardware and software since I was in my early teens - and I made tons of (innocent) mistakes.

When you get to be an adult you then know how to approach complex systems where you might not have this much margin.

Much of it is heuristics. Offloading heuristics (despite biases) is a VERY BAD IDEA.

This is an important read about commitment and identity with respect to the work you do. Although not everyone will have underlying illnesses the fallout can be as severe.

Looking back at my academic career I see many parallels in the "costs" mentioned:

- sold my twenties to institutes/labs
- traveled too much
- gave too much of myself away (for free)
- strain on mental health

https://kennethreitz.org/essays/2026-03-18-open_source_gave_me_everything_until_i_had_nothing_left_to_give

#academicchatter #academia

Man threatening to leave NATO, and invade member states, demands NATO reciprocity in a war he started. 🤡

The Onion couldn't have done a better job.

You mean the same AI that sends people spiraling into psychosis and or other wild delusions? 👌

https://www.theguardian.com/technology/ng-interactive/2026/mar/03/schools-student-ai-counselor

RE: https://mastodon.social/@Climatehistories/116092325526248754

A new pervasive world view in which people are (even further) downstream from the system is crystalizing. Troubling observations by @Climatehistories

"Small-scale (under £5,000), fast-turnaround, one-off projects are commonly sidelined by university contract offices because they represent too small an income for strained institutional resources, creating space for contract-research organizations and technical consultancies with the agility to capture this demand."

Stay out of my lane, will you. I've carved this niche.

https://www.nature.com/articles/d41586-025-04104-2

So, now it is cats, dogs and ... school busses. Do these drive on GenAI, because this feels a bit too stochastic.

https://www.engadget.com/transportation/waymos-robotaxi-fleet-is-being-recalled-again-this-time-for-failing-to-stop-for-school-buses-190222243.html

@thomasfuchs Won't change a thing IMO, as you sign away those rights with an EULA already (see ChatGPTs - most probably clicked through something similar for their search).

I think the lack of a direct line between "advice" or even the pretense of providing accurate information and negative consequences is their cop-out and not easily governed. It straddles the line with free speech etc.

@thomasfuchs True, but in the past the decision to filter out this information fell onto the user as well.

Question is can you monetize a bullshit machine, which sometimes is correct (or requires expert knowledge to leverage safely)?

The generalist nature of these tools is in part the danger and the trouble with policy surrounding it. But I guess summaries can safely be tossed. Little value there.