Jim Lai

Just a reminder that absolute resource use by AI datacenters deflects from the Econ 101 basic truth that marginal use drives pricing, and both water and electricity supply is always localized.

Who else is clocking that British PMs are regenerating faster than Doctor Who?

Of course the averaged writing voice is not a personal writing voice. And if a person never writes for themselves, they will never develop one. theconversation.com/college-stud...

Not only did Dawkins entertain AI consciousness based on chat alone, he imposed AI gender, therefore engaging with a transfeminine AI. gizmodo.com/the-father-o...

Grok is cooked, at least for a while, which will curtail the ability of the Pentagon to leverage the least hinged major US AI. gizmodo.com/elon-musk-is...

It's worth remembering that AI is more than the US companies that dominate Western media discourse. restofworld.org/2025/kenya-c...

Once again, the secret ingredient is crime. torrentfreak.com/nvidia-conta...

A power move. techcrunch.com/2026/01/05/h...

I cannot wait for the first wave of generative AI fiction to be peppered with inline product promotion, violating genre conventions and embarrassing putative authors. www.tomshardware.com/tech-industr...

12 shades of slop. gizmodo.com/ai-image-gen...

It's not a good look when even AI papers are submitted with hallucinated references. gptzero.me/news/iclr-20...

Grifters are getting out of hand with AI. Misinformation deepfakes for unethical profit. www.theguardian.com/society/2025...

Imagine Diogenes carting in a humanoid robot and having it walk on stage whlie engaging as an LLM chatbot. "Behold, a man!"

It never fails to amuse me whenever more evidence turns up that the genetic code is more variations on a theme than a fixed thing. cen.acs.org/biological-c...

It is inevitable under capitalism that in a loneliness epidemic, AI companies will see opportunity in profiting from sustained loneliness. In this era of revenue stream maximization, the best customer is a recurring customer.

Time to find out what the risks are to the ecosystem before anyone seriously tries to launch orbital AI datacenters or factories. The precautionary principle demands we look. news.climate.columbia.edu/2025/03/04/r...

My "biological name" is H. Sapiens. Binomial nomenclature is internationally recognized.

Due to a lack of data, we have been spared future attempts to train AI to one-shot a movie. petapixel.com/2025/09/23/m...

AI can now make mistakes in military planning faster than humanly possible. breakingdefense.com/2025/09/air-...

Rapture? A human body literally being raised into the sky indefinitely is not headed for Heaven. First hypoxia, then anoxia, followed by unconsciousness. Past the Armstrong limit, water in the body boils off. The end result after ebullism is a radiation-fried bloated mummy headed for deep space.

Who else is astonished at the number of people who didn't know that acetaminophen is generic Tylenol?

With AI, we've literally automated Orwell's "duckspeak". The formulas may be more complex, but that fact that we can now fairly easily detect AI slop is sufficient proof.

A recent piece about people's experiences with AI-enabled psychosis. I found it interesting that Gemini helped one man realize ChatGPT was supporting his delusions. ici.radio-canada.ca/rci/en/news/...

An ongoing trend not typically covered is how the Russia-Ukraine conflict has escalated to drone war arms race, now augmented with AI, resulting in rapid development of Skynet-adjacent technologies. Societies are sleepwalking into this future. www.defensenews.com/industry/202...

Christian Evangelicals and TESCREAL are not aligned. Time will tell how this internal conflict plays out, but there is increasing friction. arstechnica.com/tech-policy/...

Long story short, now it's been quantitatively shown that AI-generated stories are derivative slop. www.pnas.org/doi/10.1073/...

What if the danger is that AI is too anthropomorphic to trigger a sense of the uncanny valley? news.westernu.ca/2025/08/dang...

TESCREAL is killing the USA in an all-or-nothing bet on ASI.

AI is eating jobs. The only question is to what degree. www.tomshardware.com/tech-industr...

A reminder that AI is still deep into a hype cycle, and is in need of a reality check. www.tomshardware.com/tech-industr...

Unit testing AI isn't good enough in practice, according to this study. ai-frontiers.org/articles/how...

I told someone back in 2020 that everything being done against protesters then was a trial run. My only regret is not making that statement more boldly. www.npr.org/2020/07/17/8...

I'm uneasy that millions of books were destroyed in the process of feeding this AI effort. Have any works been lost to future human readers? arstechnica.com/ai/2025/06/a...

That way madness lies. Now with AI. www.tomshardware.com/tech-industr...

Bluesky is "dying" because blocking is highly effective against trolls. That cuts short a voluminous yet highly unproductive form of engagement.

The term "sigma male" is ironically apt because Six Sigma is renowned for process bloat and resulting inefficiency.

What if it's not that degrees are being overproduced, but that the oligarchic market is under-utilizing them? Talent is being squandered in the name of efficiencies and quarterly performance. futurism.com/computer-sci...

Past a point, automating one's life decisions away is servitude. LLMs have default ideologies thanks to pretraining and can be fine-tuned to be anywhere on the spectrum in principle. www.pcgamer.com/software/ai/...

We really need some way to inoculate people against using sycophantic AI like an oracle. At this rate, AI-accelerated confirmation bias will eventually turn deadly for someone. wccftech.com/after-12-yea...

Welcome to hell! One can speculate why Zuck thinks this is a good idea. futurism.com/zuckerberg-l...

The more code made by generative AI, the more stochastic vulnerabilities. I remain skeptical. arstechnica.com/security/202...

Prediction: I'm expecting a US-India trade deal to be signed later this year as the US administration seeks quick wins to deliver to its base, but it will underwhelm Wall Street. India can't replace China for manufacturing or REEs.

I've known for a while that roleplaying is a way to get an LLM to bypass its safety training. In fact, the canonical assistant role is itself a roleplaying prompt. "You are a helpful assistant." hiddenlayer.com/innovation-h...

The US executive order to push AI in schools without concern about students having access to any devices capable of interacting with AI reeks of magical thinking. This is guaranteed to turn out to be another technocratic bungle like No Child Left Behind.

Seeing headlines like this make me wonder how much of a Pandora's Box a gaming PC is becoming. www.tomshardware.com/tech-industr...

If this is the cusp of AI-powered surveillance state technology, nothing is preventing foreign actors from building AI-powered spies using the same approach. wccftech.com/ai-personas-...

We can now automate maddening bureaucracy where fake policy is made up on the spot. arstechnica.com/ai/2025/04/c...

It turns out that accurate AI is dependent upon human-generated mainstream media news. It wouldn't take much to turn AI into an amplifier for yellow journalism and state-adjacent propaganda. reutersinstitute.politics.ox.ac.uk/news/groks-r...