flir

32 ms·

571.
by flir 2y ago
Then you are not a free speech absolutist, and reasonable people will disagree about where the tipping point is. Fire in a crowded theatre? CP? Threats of violence? Hate speech?
572.
by flir 2y ago
LLMs aren't the ones making the decision to use the copyrighted information as training data, and it's that decision that is at issue here.
573.
by flir 2y ago
So to be clear, you're arguing this one specific use (machine learning) should be knocked on the head? And not all novel uses? Because "content owners to say what can and cannot be done with their data" is quite broad.
574.
by flir 2y ago
Heh. This is not the month to be making that argument. I like having food hygiene standards - it means I don't have to worry about chalk in my bread, arsenic in my sweets, or antibiotics in my beef. I honestly believe we'd be be
575.
by flir 2y ago
You're taking an "everything not permitted is forbidden" approach, which contradicts the common law principle of residual freedom. This would automatically outlaw any new use of information (eg music sampling) by default. If
576.
by flir 2y ago
Interesting. I wouldn't be able to keep that straight in my head - I wouldn't be able to predict all the low-level functions I needed until I'd built the higher-level functions. In fact I'd approach it exactly backwards
577.
by flir 2y ago
I don't believe you believe this is about transparency and auditing. You're sealioning. Where are the forensic accountants? Who uses CompSci-track college kids to audit billion-dollar orgs?
578.
by flir 2y ago
I'm a big fan of Electoral Calculus' Tribes model: https://www.electoralcalculus.co.uk/pol3d_main.html Bearing in mind all models are wrong but some models are useful, I think that one's more useful than left
579.
by flir 2y ago
Doesn't have to be an SVG, just has to be text. Pornography involving children is Right Out, regardless of the medium, and that's not going to change. The distinction between illegal porn, models that are fine tuned to write illeg
580.
by flir 2y ago
> testing powerful AI models for national security risks Er... and then what?
581.
by flir 2y ago
Mike Corley is a name some ancient Brits might remember. I just checked and his website's gone. I hope he's ok - he'd be pretty old by now.
582.
by flir 2y ago
I know I was reading think pieces by the mid-90s that accused these early adopters of being unreasonable elitists, who were gatekeeping because they just happened to get there first, and should shut up and sit down instead of complaining ab
583.
by flir 2y ago
> constantly clean-rebuilding your whole (doomed) project, gives you both some downtime and a sense of forward momentum ouch. You've thought about this, haven't you? Your ideas are intriguing to me, and I wish to subscribe to y
584.
by flir 2y ago
I didn't know there was anything that predated Core War (1984). Cool, thanks.
585.
by flir 2y ago
Most likely you'd get stuck on a local maximum. Or rather, you couldn't prove you weren't stuck on a local maximum. It's still very cool, though. Lisp is a preferred language for this because of the whole code-as-data
586.
by flir 2y ago
I found it better at reasoning, worse at coding. Not doubting your experience, just thinking how subjective it all is.
587.
by flir 2y ago
> Do we have a way to tell if one model is smarter than another at that point? Ask them to design a ranking mechanism for you. They are superhuman, after all. (I really don't think we're going to have to worry about this).
588.
by flir 2y ago
Chances of me putting a computer screen between my eyes and a lathe are exactly zero.
589.
by flir 2y ago
> This sort of injury, with the missing sensory input, triggers brain plasticity Huh. Charles Bonnet syndrome?
590.
by flir 2y ago
I'm not worried about the data, and even losing the servers would be a hiccup. But whatever site is at www.wikipedia.org is, in the minds of the general public, Wikipedia. Domain names have been seized in the past.
591.
by flir 2y ago
> Japan found over 200k ghost centenarians in their database Weeeeeell.... https://cnrs.hal.science/hal-03478634/document > Why? 150. I already said this. The children of a 150-year-old would probably be in their
592.
by flir 2y ago
That makes me wonder what the Wikimedia Foundation's disaster recovery plan looks like. The crucial bit of infrastructure (and the most vulnerable point right now) might be the domain name.
593.
by flir 2y ago
I'm sure the US system has a few cases of dead claimants. I mean, it's a country of 330m people. Dead and 150? Bit harder to believe. It's an extraordinary claim, if you get my drift.
594.
by flir 2y ago
At least we'll get a good example for the Chesterton's Fence wiki page.
595.
by flir 2y ago
Colour blindness (to take a single issue) is commonly quoted at about 4.5% globally (exact percentage varies by ethnicity so it varies by country). Your "0.5%" figure wildly underestimates the issue IMO. I agree that local setting
596.
by flir 2y ago
In the book that talk is based on (I think it might be The Blind Watchmaker?) he says (paraphrasing) that he's using the example to illustrate a process, and the exact start and end points don't really matter. If you prefer an ear
597.
by flir 2y ago
I think of features. Code can always be fixed (in theory, at least: the Arrow of Software Development rarely points away from entropy), but "we're gonna strip this feature set back to basics" is even rarer.
598.
by flir 2y ago
Ever heard the expression "bad money drives out good"? Free content drives out paid-for content. It would only take a tiny percentage of sites going ad-supported. All the users would rush to them, the microtransaction sites would
599.
by flir 2y ago
I wouldn't go full-on conspiracy, because I expect the impetus came from the UK, but... I doubt it would have gotten this far without tacit US gov support.
600.
by flir 2y ago
Mate, the "take all steps necessary to root out paedophiles" referendum would be won with 95% of the vote. It's all in the framing, you know that.

More ›