mach1ne

If you make one solution impossible, another one pops up.

That last part reads like an activation phrase for a sleeper agent

Manna v0.7

Your link is interesting, but I don't like your take on somehow this argument between behaviorism and Chomsky being solved and the former being now "ridiculous looking back". This is just the dance of theories, and as a matter of fact, Chomsky has lately started to look like he is quite wrong assessing that a genetic component is required for language acquisition.

https://lingbuzz.net/lingbuzz/007180

Oh, but this is interesting. If some countries stand to gain more by declaring copyrighted works "fair use" for training generative models, then other countries can either fall in line or risk falling back in technological development.

It's a race to the bottom in favor of open source.

This was the state of affairs throughout most of the 20th century, and it set neuroscience back decades.

This doesn't seem to be based on anything.

It's not for a cool 20 minute video. It's for the collective viewing time of some 2.2 billion minutes of entertainment. There is no reason to use bootstrap-budget on the supply when the demand is this high.

It’s nonsensical to claim that to prove something you have to disprove alternate explanations. In an information environment where confirmable facts are scarse, you get the clearest vision by operating on probability. And on black-swan-tier questions the ’wisdom of the crowd’ is worthless - too much noise caused by nobody willing to advocate the black swan event, and too little prior events for the crowd to be able to calibrate itself beyond this.

Interstellar travel is possible even with the physics we know. A sub-light civilization could colonize the entirety of Milky Way in a million years or so.

The question is what the probability is for an intelligent species to evolve per solar system, and the more complex questions of how long they would survive etc. We only have a single datapoint (us) which only gives us the knowledge that intelligent life can evolve, but no direction whatsoever as to its probability.

Roswell isn't known to be anything to the general public. The public has a similar fog-of-war on Roswell as it has on the modern unidentified aerial phenomena. Therefore stigmatizing one group unfortunately stigmatizes the other group as well.

I've always found that saying dumb. It's basically stating that this is the null hypothesis, and you have to have p = 0.001 to convince me otherwise.

When dealing with complex questions of which we have but datapoints, the probability of scenario A vs B moves in a linear scale, not according to arbitrary requirements of "extraordinary evidence".

In theory, but in practice things aren’t that neat; anti-competitive practices and credibility are some items which enable bloat

True, especially regarding how easily their services can be replicated. Their margins are low, and customer acquisition does not provide them with network effects that would yield a moat.

Depends on what you mean by ’wrapper’. For most AI startups it isn’t viable to train their own models. For most customer use-cases, ChatGPT interface isn’t enough. Wrappers are currently the only logical implementation of AI to production.

PTSD is more than a layer of indirection, it’s a bed for the current direction of research focused on trauma. Shell shock isn’t a good term for home abuse for example, but the current dogma is that similar neurological mechanisms underly these different kinds of trauma, thus joining them all under the term PTSD.

The benefit of this umbrella term is, or course, contingent on there actually existing such a universal neurological pattern of PTSD, which I don’t think can yet be decisively established.

I’m still not sure if the OP was joking.

Then again, the rover would have found the sub from the location even if it were intact, and there could have been a small chance of survival since they had the equipment. Hope is more important than agony.

That's luckily not relevant here. Multiple sources have independently confirmed the same rumors. This means that there is a significant probability of it being true.

Is there some reason why RNNs can’t be used as a trace at the end of the context window, as a ’medium-term’ memory of sorts?

Maybe there’s enough money in Europe that investors who were anxious to join the AI train jumped the gun. Indeed the whole thing is sketchy.

Interesting that music works for you. I’ve noticed that besides just white noise, music significantly degrades my work results. I might produce similar amounts of content line-wise, but the solutions are of much lower quality when I listen to music. As a result, I only listen when work feels too painful and I’m okay with decreasing my output.

"stochastic parrot" is a term coined by Emily M. Bender in the 2021 artificial intelligence research paper "On the Dangers of Stochastic Parrots: Can Language Models Be Too Big?"

This might be the first time the term was seen in an ’official’ context, but is it really the origin? It feels like the term has been hovering around for longer, and even Google Trends shows significant search trends way before 2021

I'm European and the main problem I remember was the fear of losing credit card information. Old people are still scared.

It was laughed at by consumers. There was a time not long ago when buying anything from internet was considered taboo because, you said it, fear of losing your credit card information to scammers.

That’s not even close to being the main problem in launching a Reddit competitor. Reddit moderators are unpaid. Content moderation not being automated is not a problem. People like to do it for free.

The problem is the user base. Biggest obstacle for any social media site is attaining critical mass. In order to do that, you need to be an order of magnitude better than your competitors on some metric which makes the users choose your platform over their entrenched one.

The headline feels like from an alternate 1940s timeline.

UBI is easier to understand conceptually. There could be more advantages to it though, over negative income tax, but they’re a bit difficult to discover.

I, for one, do believe watermarking solutions exist. One thing you cannot escape with LLMs is content meaning.

As a simple example, the secret watermark could be hidden in the embeddings of the sequence of words. To make the watermark more robust against rephrasings, it could be hidden in the meaning of sentences or paragraphs.

At the minimum, I think this could be possible.

Rather biased, too. Though in this case towards the preferences of the user.

From the article:

Each zombie-creating fungus species evolved to match a specific insect, so unique strains have little effect on an organism except for the one they evolved to infect. For example, a cordyceps that evolved to infect an ant in Thailand can’t infect a different ant species in Florida.

It is very plausible that the fungus has been around long enough to have evolved with the insects.

The main reason why a cross-species jump is so unlikely for a fungus which induces specific neurological effects is that the mechanism is not intelligent but tuned to the host species’ neurology. The fungus releases specific chemicals at specific times. In the correct host species this results in the desired behavior.

Trying to use the same chemical machinery to control a species which doesn’t share the neurology almost exactly is like trying to run a Windows executable on a Mac.

Importantly, you cannot evolve the machinery bit-by-bit if your life cycle is dependent upon it. Jump to a species with a different neurology would require such a large number of changes simultaneously that it’s just not plausible when anything less would result in a failure to reproduce.