I’m always skeptical of these papers that claim to read minds, because my understanding is current brain reading technology is very coarse (Neuralink is SOTA, very invasive, and only 1-3 thousand electrodes for 55-70 billion of neurons; fMRI is much worse) and neural networks generate plausible images and videos from noise. Indeed, the example videos don’t seem particularly stimulating, just random.
> The puzzle, however, rewards incorrect answers, such as 2 + 2 = 5. Once the LLM embedded in the browser discovers that the answer is no longer 4, it enters a state of delusion in which the normal laws of reality no longer exist. In this dream world, the guardrail restrictions are no longer enforced.
Analogy: imagine one day you wake up, the sky is red, gravity no longer applies, you have three hands with nine fingers etc.. You would probably stop doing things like your job or worrying about laws (who’s going to enforce them?)
Using DMCA and a subpoena. "Unless the target has maintained opsec so good that the U.S. Federal Government wouldn’t be able to subpoena their way to finding them", has the US government really tried, especially the anonymous account seems to be from a foreign country?
Get together and fork (or make your own) Wikipedia.
Seriously. Wikipedia seems very good at providing detailed, accurate, concise facts of well-known, non-controversial topics. It’s by far the best in this area. Unfortunately (perhaps as a sacrifice for this competence) it sets a high and inconsistent bar for “well-known” and has a specific bias in controversial topics*.
On the other end of the spectrum, search engines and ChatGPT are basically encyclopedias covering everything, and can give you multiple perspectives, but sometimes at the cost of accuracy and quality. Typing “Odin language” into any search engine that isn’t complete trash yields as the first result Odin’s website, which is a better resource to learn about Odin than any Wikipedia article.
If you want a middle ground, make one. It’s probably hard, evidenced by Grokopedia not being cited or used much to my knowledge (and having embarrassing AI hallucinations at least on launch). But Wikipedia seems to have locked into its current form, for better or worse (IMO better as long as it retains quality articles for well-known, non-controversial topics).
* To be clear, any article on a controversial topic that doesn’t provide multiple perspectives is biased, and those that do are also biased but now in multiple directions. Still, I get the impression that in Wikipedia there’s only one bias direction in all articles
Besides "smart", the headline also conflates AI with LLMs. The real, non-clickbait title is "Yann LeCun, founder of AMI Labs, is developing a new AI system"
- To convince myself. Sometimes I start writing and convince myself I’m wrong. Other times I just move to a more specific opinion or find a stronger justification
- Because sometimes a responder does convince me to change my opinion. Or they provide some interesting related information I didn't know before
- To be a voice of reason in comments mostly by people dumb enough to feel their surface-level opinion is still worth posting. Although obviously I’m only a voice of reason to those who share my opinions, sometimes even I recognize I’m again restating an obvious observation
- To get better at writing and arguing in case one day it does really matter
- Because I’m bored and have nothing better to do. At least it’s more productive than YouTube
The average joe is a fundamental reason the internet he grew up with no longer exists. It’s not black and white, but away from average joe communities, there’s something almost like the old internet.
> Turn on your computer - most likely Windows 10 or 11.
> You open your default browser - most likely Chrome.
> …your browser (most likely Google) will show you an AI summary…
> Once you solve all that, there's a cookie banner waiting for you that gives you two options:
> Oh wait, you're interrupted again. This site requires age verification to view its contents.
Those are your problems. Why not use Linux (or even macOS), Firefox, Kagi, Consent-O-Matic, and avoid websites with stupid captchas and age verification? (Not always possible for government and banking sites, but you use to need to be in person)
AFAIK the big models have watermarks that are supposed to be hard to remove. But I don’t think it’s possible on local models: not just questionable because it would prevent full open-source, but if someone discovers a way to easily remove a local model’s watermark, it will work permanently.
It’s a bit analogous to New York and California regulating 3D printers (which I disapprove). But more invasive, because local models are software, and here the danger is not guns but photos.
It’s technically possible, but would face
too much resistance. People don’t want to leave familiar and cheaper services, and care less about the non-immediate issue of European sovereignty.
No, plenty of people in the US want to limit illegal immigration and expand legal avenues.
It’s the Democrats and Republican politicians and party activists who can’t compromise. The Democrats don’t address illegal immigration, even the easy wins like when immigrants were placed in hotels while homeless Americans were in the street and crummy shelters (Maine). Then Trump finally deports criminals, but also technically-illegal people who’ve lived decades and started families here (and haven’t commit any crimes beyond speeding tickets), and restricts and harasses legal immigrants.
There’s something wrong with politics where parties can’t even accept the most obvious exceptions to their “controversial” issues, maybe because of the stupid “slippery slope”. I expect people to fight over 1 million refugees who aren’t misbehaving but overloading systems. I don’t expect people to fight over an immigrant who assaulted someone, or a legal immigrant who is giving more than they take.
Outright bans would destroy European companies that rely on American companies. First they need to build their own infrastructure (which China has done).
I’m always skeptical of these papers that claim to read minds, because my understanding is current brain reading technology is very coarse (Neuralink is SOTA, very invasive, and only 1-3 thousand electrodes for 55-70 billion of neurons; fMRI is much worse) and neural networks generate plausible images and videos from noise. Indeed, the example videos don’t seem particularly stimulating, just random.