Prometheus

I agree, but the counter argument is that so far the biggest improvements have simply been bigger models, not more structured models. There was a more subtle point at the end of the article suggesting that we shouldn't be trying to inject human knowledge directly into AI because the mind is complex beyond our understanding and if we don't know how knowledge is codified in the brain, we don't have a basis to codify it in AI. I don't understand what you're trying to say.

I gave this example early on in the thread; it raised questions in MMA. I agree that Curious laymen was unfairly set upon in this thread.

I learnt classical stats so i'm in no position to defend Bayesian approaches, but i'm not sure we need to dump Bayesian techniques. They are both tools and some jobs may be better suited to one or the other. I know there are a lot of physicists on this site so i wonder what they would make of a Bayesian approach to something like quantum theory. Also, something i've never understood: both approaches are consistent with the Kolmogorov axioms, so in what sense are they really different at a fundamental level? I'm surprised this topic hasn't reared its head on this forum before.

The biggest lesson that can be read from 70 years of AI research is that general methods that leverage computation are ultimately the most effective, and by a large margin. According to Deepmind's Richard Sutton the bitterness is that injecting human knowledge into an AI systems results only in short term gains and that general statistical methods that leverage computational power perform much better in the the mid to long term. GPT-3 has 175 billion parameters and doesn't seem to be approaching the limit of what a simply bigger model can do. Some speculate GPT-4 will have trillions of parameters. A debate between Yann Lecun and Gary Marcus shows there is a split in the community regarding the wisdom of this approach. I wonder whether there is a hard limit to how much computation that can be thrown into a system. Some say we are already coming up to the limit of Moore's law, and that parallelisation has its limit in Amdahl's law. We also know that the human brain doesn't need to consume nearly as many resources to achieve far more complex behaviour than current AI (based on the relative energy efficiency of brains vs computers). We also have reason to want to model the human brain with AI to deepen our understanding of the mind. And if we are to create AGI, presumably it will easier to align our values to an agent if it is in some sense based on human biology (although Sutton's argument would be that actually, more computation will achieve that end). My instinct is that at some point the vast complexity of the world will require that AI systems parse things down into simpler units of understanding in order for them to navigate it. Does anyone have any opinions one way or another?

So we have all these ones and zeros in a superposition of states, but for the final output we want it to collapse to a single state: the right answer. How is this collapse controlled to that end? Or is this not how it works.

As far as i can tell from the BMJ editorials related to those studies, the concern in sports medicine seems to focus mostly on the nuances of competition rules. For instance, there are calls to make the period of hormone therapy longer before allowing transwomen to compete with ciswomen. There have also safety concerns raised in the MMA community, though i can't find anything related to that with my glance at the academic literature. Haven't time for a deeper dive so i'll have to leave it there.

I came across this review in the Journal of Medical Ethics which suggests there is enough evidence that tranwomen athletes have an unfair advantage under current IOC guidleines to warrant discussion and further research in the area. This later review in Sports Medicine came to the same conclusion.

Here's a real example from MMA. I don't know enough about the issue to comment, but i can see why people would be concerned.

Here's a good first guess as to what happened:

Elon gave SN10 a 60% chance of surviving, so the magic was just in landing. They likely would have just dismantled it had it survived, but they had a rapid unscheduled disassembly. Instead it's a great opportunity to explain SpaceX's development model - rapid prototype production and testing, push it to its limits (including blowing it up), learn as much as possible, build the next prototype, repeat until it works. It'll be interesting to see if starship makes it to the moon before NASA's SLS (both aiming for 2023). P.S. I was watching a live feed - they mentioned that NASA had a chopper outside an 8 mile exclusion zone to take video.

Are you sure 'religion' opposes stem cell research. Or by religion do you mean certain subsets of some religions - particularly those vocal in America?

What did they perform PCA on? Presumably some kind of genetic dataset? The problem with this kind of dataset is you can similarly label jockeys and basketball players on a and get a visually appealing separation. Does this mean there is a difference between these two types of sports people such that you want to call them different races? You could. What is 'biologically real' is to an extent based on how we choose to categorise things - apparently even the concept of species gets murky at some point. So i guess the question is whether treating race as 'biologically real' is useful. Race and ethnicity are still used in medicine, albeit badly (often no distinction is made between the two). They're used as proxy markers for what is really pertinent, which changes depending on the context. Sometimes it helps get an idea of diet and exercise habits, though you would be better off directly measuring them if possible (surprisingly difficult), and sometimes you want an idea of genetic propensities, but again you would be better off measuring those directly now that is possible. However, genetic testing is still expensive. Hence the double risk of using race in medicine is that it can lead to a conflation of genetic factors with cultural behavioural factors, and it is a poor substitute for direct genetic testing.

Watch some med school lectures.

Almost anything, including an algorithm, can be considered a blackbox. The defining characteristic seems to be something viewed in terms of inputs and outputs, without knowledge of the internal workings. A disease classifier (i.e. an algorithm) would be considered a blackbox to a medical professional who doesn't think in terms of hyperplanes and decision boundaries.

I only dabble in this, but my understanding is that you want the manuscript to be as polished as you can make it before sending out to an agent. That agent may help with some editing, depending on their experience and expertise, before sending it to an editor. If accepted you will then work with the editor in getting it into final shape. If you have the money you can pay for an editor to look at your manuscript (for instance, this is recommended for people who take self-publishing seriously). Way before all that you might want to send the manuscript to beta-readers - people you trust to look at early drafts to give advice and feedback. Not friends and family - unless they actually give objective advice - but usually people you know in writing circles. Are you writing sci-fi by any chance?

Just to add another perspective, neural networks are being used to model increasingly complex phenomena and in general the more parameters the model has, the better they perform - so long as it has sufficient data on which to learn. I mean, most of them have more hyper-parameters than normal models do parameters.

Any model selection process which uses AIC or BIC (or any other variable penalisation method) is explicitly using Ockham's razor to prevent over-fitting of models. I'd wager their use is increasing not decreasing.

I agree it's not something we should obsess over, but it's certainly something worth thinking about. As the stats show, there a relatively few sudden deaths so chances are death is something you will experience: the last thing. By spending some time thinking about the subject we can improve that experience In my experience most people are woefully under prepared for that final experience..

Are people more scared of crashes and burning than disease? Even if it's true of some people i suspect it's heavily dependent upon age and what's happening in your part of the world.

I got the same numbers. This discovery was considered noteworthy because they were looking at equatorial regions were it was expected no water would be present. Presumably there is a greater abundance at the poles? Couldn't find reliable figures on a quick search. Interesting question about how much is needed to enable colonisation. The proximity of the moon (relative to other celestial bodies) means any colony need not be fully self-sufficient. Speculating, if He3 became a sufficiently valuable commodity then companies would be willing to ship all the water they needed up from Earth - the tricky word being sufficiently. In short the viability of a lunar colony would be dependant upon economic factors in addition to the engineering ones.

That's a problem but is this legislation going to change that - from what i understand the Aussies are talking about diverting ad revenues to mainstream media? I think the fundamental problem with online echo-chambers is that we have algorithms optimise only for traffic interacting with the reptilian parts of our brains which are the part of us most easily triggered to smash that like button, as the kids say.

But are they complicit or is it just changing the ways people interact with media, thus making traditional routes less viable? It reminds me of high street chains complaining about internet retailers: the former have to pay taxes on property, the latter not so much, giving internet based operations a competitive edge. But, in this example, nothing nefarious is being done by the internet based retailers. Maybe the solution is to innovate rather than legislate?

Banning any websites will encourage more people to use VPNs. If a sizeable proportion of a population are using these then any government loses any ability to control their internet.

If you're thinking beyond low-Earth orbit there's also the interaction of solar activity and the Van Allen belts, which can be influenced by more subtle solar activity than CMEs, as well as various other factors. I believe Apollo just burned through it as quick as possible to minimise exposure, but if we ever get to a point where humans frequently traverse this zone it might be taken into account. But that's not a dramatic effect like you're envisioning, just an increased lifetime risk of some diseases.