ingoboltz.bsky.social - Profile | ThreadSky | a Reddit-style client for Bluesky

Poe just cut the price of it's entry level subscription in half. Probably reflects the DeekSeek effect (cheaper-to-run models) but also consumer demand that's not quite "skyrocketing" as the model vendors would want you to believe... x.com/poe_platform/s...

submitted 5 days ago • 0 comments

Sabine Hossenfelder was never a destructive critic of LLMs. But here she's saying what so many of us who have no vested interest in hyping them experience every day.... x.com/skdh/status/19...

submitted 5 days ago • 0 comments

High profile revelation of a company promising AI agents that can replace your bothersome, expensive human employees, only they didn't work as promised and customers walked. So 11x just faked praise and inflated revenue. Outlier or just the first report? techcrunch.com/2025/...

submitted 6 days ago • 0 comments

Hopefully something useful coming out of current epic AI spending: Better weather forecasts! :-) Project Aardvark: reimagining AI weather prediction www.turing.ac.uk/blo...

submitted 6 days ago • 0 comments

Ed Zitron laying into lazy AI industry reporting - worth reading, if with the qualification that Ed should really get an editor to help condense his rants. :-P But the essence in there is spot on. The Phony Comforts of AI Optimism www.wheresyoured.at/...

submitted 7 days ago • 0 comments

Non-technical Sunday post, extracted from "Left is not woke" by Susan Neiman: "I am not an ally. Convictions play a minor role in alliances, which is why they are often short. If my self-interest happens to align with yours, for a moment, we could form an alliance. [...] 1/4

submitted 8 days ago • 1 comment

Enjoyed reading this! Pure Independence collabfund.com/blog/...

submitted 8 days ago • 0 comments

Don't want your content be scraped by AI model maker bots? Since many bots ignore robots.txt until now there was little you could do. Now, Cloudflare prevents theft with custom-made AI slop. Gotta love it... www.theregister.com/...

submitted 9 days ago • 0 comments

We were just talking about the re-emergence of "parasocial relationships" were we not? There you go, we have the AI incarnation of the concept. And I'm not sure at all that it's "neutral"... Did Google Test an Experimental AI on Kids, With Tragic Results? futurism.com/charact...

submitted 9 days ago • 0 comments

In today's misinformation news: a new Trump audio fake that was widely shared, and an example of the tendency I posted about last week to use visual templates of reputable news media (here Al Jazeera) to lend credibility to video deepfakes. www.newsguardreality...

submitted 9 days ago • 0 comments

I suspect we'll hear a lot more about "parasocial relationships" in the near future. The concept flared up last in the context of social media, but with "AI companions" on the rise, will people soon form "parasocial parahuman" relationships with LLMs? :-/ www.verywellmind.com...

submitted 10 days ago • 0 comments

This should be interesting to watch! If they do run on par with humans, we should give the robots their own marathons, no? What would a good, motivating prize be for a winning robot? :-P www.scmp.com/news/ch...

submitted 10 days ago • 0 comments

After the French scientist who was just turned away at the US border for being critical of Donald Trump I am tempted to share a non-AI text on the gradual erosion of democracy. I grew up in East Germany; the new authoritarianism in the west worries me! www.codastory.com/re...

submitted 10 days ago • 0 comments

One always suspected that big LLM makers would have had to use pirated material to get to the volume of training data at the cost (free?) and speed (now!) they wanted. Having it on record in court now will show if "Big AI" really is legally untouchable. www.theatlantic.com/...

submitted 10 days ago • 0 comments

For a change, a positive AI story! Nobody's job eliminated, more people enjoying theater, and at accessible cost... LLMs contributing something good to society! :-) www.semafor.com/arti...

submitted 10 days ago • 0 comments

No Semafor, insisting on user privacy is not Apple's "core problem" with "new SIRI". It's that they buckled and over-promised on what their LLMs can do because their competitors constantly do the same (and get away with it because of fuzzy benchmarks). www.semafor.com/news...

submitted 11 days ago • 0 comments

Non-programmer employees actually trying to use the currently available LLM tools in their dayt-to-day work know that they underwhelm. Those of their managers who aren't frequent users keep running after the hype. Still hammers looking for nails. www.axios.com/2025/0...

submitted 12 days ago • 0 comments

Some interesting AI media regulation news from Europe. It will be interesting to see whether these strong provisions can actually be enforced and will lead to less "AI slop" in Spain. Spain to impose massive fines for not labelling AI-generated content www.reuters.com/tech...

submitted 13 days ago • 0 comments

Insightful rant on how even Apple, once the symbol of product design and the anti-thesis of vaporware, caved to market pressure and promised AI features that they haven't made to work (and maybe never will). Something Is Rotten in the State of Cupertino daringfireball.net/2...

submitted 13 days ago • 0 comments

This. Nothing to add. OpenAI brings you statistically average literary fiction pivot-to-ai.com/2025...

submitted 13 days ago • 0 comments

Worthwhile, in-depth study on LLMs failing to correctly attribute answers retrieved from web searches. Considering that "LLMs as search engines" threaten to kill the web as-we-know-it by cutting out content producers from revenues, this may be... good? www.cjr.org/tow_cent...

submitted 13 days ago • 0 comments

If you tune-out people with vested interest in the continuation of the LLM bubble, and listen to technical AI experts instead, this is what you hear: www.newscientist.com...

submitted 14 days ago • 0 comments

Beautifully summarized! OpenAI asks the US government for the moon on a stick pivot-to-ai.com/2025...

submitted 18 days ago • 0 comments

Apologies to non-German speakers! :-P Parteispenden deckeln: Jetzt Appell unterschreiben! In DE sind unbegrenzte Spenden an Parteien erlaubt. Doch Superreiche sollten ihren Interessen nicht mehr Gewicht verleihen dürfen. Unterzeichne hier: www.lobbycontrol.de/parteispende...

submitted 18 days ago • 0 comments

Yes, I know, I share lots of "bearish on LLMs" material, and here is another, well written one... Enjoy! :-P A Bear Case: My Predictions Regarding AI Progress — LessWrong www.lesswrong.com/po...

submitted 19 days ago • 0 comments

Last week, Apple further delayed its already announced "New (LLM) Siri" - unusual for a company famous for launching products only once they are really ready. Did pressure from the AI-hype wave make Apple announce half baked stuff to "stay in the race"? daringfireball.net/2...

submitted 20 days ago • 0 comments

If you haven't yet, try this conversational voice AI (scroll down a bit in the article). The human emotions mimicry is mind-blowing. Imagine this AI calling you to deploy misinformation or scam you... Crossing the uncanny valley of conversational voice www.sesame.com/resea...

submitted 20 days ago • 0 comments

Cheap-fakes (like grafting mainstream media logos on misinformation) still dominate, but sophisticated deepfakes like the hybrid "organic+AI" Marco Rubio Starlink video should really worry us. Bad Actors Fake Being Real News Outlets at Alarming Rate www.newsguardreality...

submitted 20 days ago • 0 comments

Using AI to infer political leanings of students who have already been issued visas. Welcome to the "1984" version of the US. www.axios.com/2025/0...

submitted 21 days ago • 0 comments

I don't know what's happening at Nature but initiatives like this should really make us downgrade its prestige by a lot. Replacing human review with LLM review, as "time-saving" as it might be, will be deadly for article quality and author motivation. pivot-to-ai.com/2025...

submitted 21 days ago • 0 comments

Some insights from FP from the trenches of a project aiming to eliminate humans with LLMs in the UK government. A Labour party run government, mind you. Wasn't labor about protecting workers at some point? The U.K. Pivot to AI Is Doomed From the Start foreignpolicy.com/20...

submitted 21 days ago • 0 comments

Here's the founder of Hugging Face illustrating why today's LLMs will not generate innovative scientific breakthroughs (as frequently promised by model makers) because such true innovation has always required stepping outside the "training data". thomwolf.io/blog/sci...

submitted 22 days ago • 0 comments

Remember the claims of hyperscalers that they were running data centers 100% on renewables, or at least rapidly advancing towards that goal? Turns out, not really. Creative accounting with RECs combined with exploding capacity build-out for AI, lovely. www.theguardian.com/...

submitted 22 days ago • 0 comments

Stated by Lucas Joppa, Microsoft's ex Chief Environmental Officer, as cited in The Silicon Valley Consensus & AI Capex (Part 1) thetechbubble.substa... (which is itself worth a read, especially the part on projections of new gas power plants based on "expected" AI demand.)

submitted 22 days ago • 0 comments

This might help explain why in my own experiments with LLMs analyzing longer texts, the results have always been underwhelming. Long-Context Embedding Models are Blind Beyond 4K Tokens jina.ai/news/long-co...

submitted 22 days ago • 0 comments

Gary Marcus calling out starry-eyed AI "reporting" by major news outlets where ALL punches are pulled. The role of the serious press cannot be to just take corporate marketing at face value; for that we have their sales brochures. We deserve better! garymarcus.substack....

submitted 24 days ago • 0 comments

You'll hear more about "LLM grooming" in the future. Groomers deploy "dark misinformation" content produced just for the crawlers that hoover up web content to train AI models, poisoning them in the process. The content never shows up on the visible web. www.newsguardreality...

submitted 26 days ago • 0 comments

Excellent little essay on anthropomorphizing AI, exemplified through a closer look at Claude 3.7's system prompt. Truly insightful. ai-cosmos.hashnode.d...

submitted 31 days ago • 0 comments

Another acidic, yet spot-on comment by Pivot-to-AI on GPT4.5 -- OpenAI themselves seem at a loss as to why you would actually use it, and by pricing the API 15-30 times more expensive than GPT-4o, they really discourage you from finding out, anyway. pivot-to-ai.com/2025...

submitted 31 days ago • 0 comments

Ignore the aggressive wording, in here are two academic studies questioning the utility of benchmarks that model makers use to claim improvements for new models they release. A model being "better" than another is still largely a thing if "vibes". pivot-to-ai.com/2025...

submitted 34 days ago • 0 comments

Not at all sure whether "agents" in the pure sense will happen anytime soon (since LLMs don't truly reason nor plan) but calling out the hype is worth it in itself. www.linkedin.com/pos...

submitted 35 days ago • 0 comments

Zuckerberg launched humanity into ever increasing echo-chamber isolation and normalized unhinged aggression in online discourse. Altman has now brought us into a new era of non-learning enstupification: New Junior Developers Can’t Actually Code nmn.gl/blog/ai-and-l...

submitted 35 days ago • 0 comments

It's because of stunts like this that I distrust breathless AI miracle stories I come across online. Google Co-Scientist AI cracks superbug problem in two days! — because it had been fed the team’s previous paper with the answer in it pivot-to-ai.com/2025...

submitted 35 days ago • 0 comments

A worthwhile non-AI read. Gangster Party www.hamiltonnolan.co...

submitted 36 days ago • 0 comments

Beautiful legacy post from 2010! :-P Google CEO Schmidt Fuels Critics With Controversial Privacy Remarks searchengineland.com...

submitted 39 days ago • 0 comments

Step-by-step guide on how to forge C2PA content credentials. Illustrative. ShmooCon and C2PA Forgeries - The Hacker Factor Blog www.hackerfactor.com...

submitted 39 days ago • 0 comments

Are AI vendors admitting that hallucinations will stay? “Accept that it is imperfect"..."Rather than say, ‘Will AI do something wrong’, say, ‘When it does something wrong, what are the operational mitigations that we’ve put in place to deal with it?’" www.wsj.com/articles...

submitted 40 days ago • 0 comments

x.com/rao2z/status/1...

submitted 41 days ago • 0 comments