At $0.17/KWhour, what will it cost to run? And keep cool?

Can it heat my house?

Consumers need to be forewarned before endorsing widespread corporate censorship and cowardice. No company should be able to manipulate media and then hide from the consequences of everyone knowing they are gutless to the bone.

If profs used AI ro both write and grade exams, as well as solicit feedback from AI and students on both tasks, maybe exams would improve. If course profs should be able to spend less time on the boring parts of teaching (esp exams and term papers) while becoming mire inventive at it. Ideally they could also use AI to improve the broader learning experience itself, since jettisoning the lecture model of teaching is decaded overdue, not to mention multiple choice and true/false tests.

I'd be very surprised if the subjects used to demonstrate the system were folks of color: hispanic, middle eastern, asian, or black. As Timnit Gebru showed in 2017, auto facial recognition performs poorly on folks whose skin isn't as reflective as caucasians. Unfortunately, they're also the folks the FBI generally want to track most. So what's the value of system that can't reliably recognize the typical targets of interest? Propaganda, most likely.

When developing new drugs, animal testing answers several questions. But this technique addresses only question #1 (and maybe #2) below.

1) What is the molecular mechanism behind the disease?

2) How does a new candidate treatment 'fix' the disease by interfering with the cause of breakage (antagonist) or by replacing a broken component (agonist)?

3) How can we measure the disease's progression, or after treatment, the healing of diseased tissues in all affected tissues?

4) Does the candidate treatment cause toxicity in other tissues, especially liver, kidney, brain, or heart? What is its cause? Can the toxicity be fixed?

Then, after all these questions are answered for these simpler models, they will have to be answered again in human clinical trials. There the answers may change significantly due to the many more interacting variables present in whole human beings that were absent in the simpler models used earlier.

Like in-vitro organoids, this is an step-wise improvement to the current need for animal testing. But it's far from a final solution.

FB's entire business is selling your name, behavior, and demographic profile to others so they can market and sell to YOU. Thus, *every* FB product contains a way for the seller to contact or track you. That identifier + personal info is what the EU demands to know, as well as who bought it and how they use it. The audit trail of accountability only *begins* with FB.

Get used to it Mark, soon there'll be no place to hide what you do to your customers and there's nothing you can do to stop it.

Like DP, a âoesmartâ algorithm like the one described that first uses a prepared (compressied) set of high likelihood matches, and then is followed by a slower pass of (uncompressed) naive search sounds pretty much like dynamic programming. The learning phase is not well described by the OP. Presumably there is some use of external information which produces new (or revised) choices to be integrated into the compressed set of choices, thereby improving future performance, adaptively and more optimally.

> Telling people that a drug is good for treating a particular condition when the evidence to back up that claim isn't there is definitely not good.

Yes, that's why drug efficacy decisions are never based only on observational studies like the one(s) mentioned here. I work for a major pharmaceutical co. and lesson #1 here is: respect the research process. It's rigorous and formal for damned good reason. There be dragons.

If you want to conclude that a drug has efficacy, the only scientifically accepted way to do this is to conduct a blinded randomly controlled trial (RCT) with a sufficient number of patients who have been selected based on a sufficiently broad distribution of biological/genetic strata. Only this has a chance to ensure the results *do* apply to most people in a population and not just a small cherry-picked subset, AND are reproducible. Until a RCT study follows this protocol, the medical community will not discuss efficacy; they will only discuss the *possibility* of efficacy and the need for a RCT to formally measure it for a) reliability (e.g. p>.05) and b) impact (e.g. mortality). Only then is a drug deemed ready to a) effectively treat and b) do minimal harm.

So yes, it *is* inappropriate to claim efficacy based only on a small set of observational cases from an uncontrolled population (if that's what the OP did). And you truly cross the line when you propose that individual clinicians experiment on sick people without rigorous external oversight by a research board. There's been WAY too much "cowboy medicine" in the past year. It does harm when patients insist on getting unproven/ineffective treatments because some yahoo has loudly served up his uninformed opinion. So I'm glad social media are trying to end this (mal)practice.

Wow. Lots of overlap with my tastes, esp. Amélie. If I may round up to 80...

---73) The Lion in Winter
---74) Interstellar
---75) A Man for All Seasons
---76) MP's The Meaning of Life
---77) 2010
---78) The Fifth Element
---79) The Bourne Trilogy
---80) The Matrix

The summary info on that site is a mess. Based on those plots, the total number of confirmed cases so far in France is about 180,000. The total number of deaths is about 26,000. Thus the rate of deaths is about 14%. That's high, but far less than the 32% you mention (which *is* on the site, but is clearly wrong).

And of course, the majority of COVID infections are not confirmed, and a very high fraction of non-hospitalized cases recover fully at home. That data is missing entirely and would bring down the death rate a great deal.

Each study treated patients at a different stage of the illness. The negative study (Lancet) treated patients in late-stage severe illness and failed to show impact. The positive study treated patients early in the illness and did show impact.

All treatments of viral infections attack the virus at different stage in its life cycle. That's why HIV therapies almost always employ drug cocktails where each of 3-5 components attack a different viral mechanism as it grows and propagates.

The implication from these two studies are that Remdesivir *is* useful in treating early stage but not late stage COVID-19.

Yes, I'd be impressed if this could identify relevant evidence to support a cogent argument. But that's not the same thing as reasoning or employing logic. They're probably just recognizing statements as being relevant (via word2vec) and then constructive vs destructive via sentiment analysis. Then they refine their ranking of candidate evidence based on the credibility of the source. For now, I suspect they probably prefer positive fare and toss the negative.

My first challenge to them: Can the system recognize evidence that disproves a thesis? Or only that which supports one? Contradiction would be harder to master than assent since it's tough to separate helpful negative comments from unhelpful positive ones (assuming they really DO toss all the unhelpful stuff by default).

And I'd LOVE to see how well this system can judge arguments and claims in the wild, like fake news. These days the interweb is in desperate need of better moderation.

Knowing what you don't know is essential to wisdom, be it artificial or natural. AFIK all modern AI systems simply guess; they never consider answering âoenone of the aboveâ, or the likelihood this question is atypical, or even whether they shouldn't answer because they're clueless.

Curiously, in the olden days of AI before deep nets, logic- and knowledge- based methods were much more common. Use of circumspection and counterfactuals were central to building models and systems that could reason and had a clue as to their limits.

Perhaps it's high time we renewed our acquaintance with Asking ourselves âoeWhat do I know?â And âoeHow well do I know it?â

What if the robot is annoying you? Can you physically push it away then? Is it OK to shout at it or act menacingly to scare it away, without making contact?

How annoying does a robot have to be before you can righteously defend yourself from it? Does the robot deserve the same respect you'd give a human, or something less? Interesting questions.

A stunning movie. I never read the book, but felt Moore was overcritical of the film. Its effect on me was jawdropping.

To set the story amid in a world that reelected Nixon to a third term and outlawed heroes; to construct a band of heroes whose values ranged from the best to the very worst of humankind; and to devise a scenario that so 'deformed' the heroic act such that it could righteously kill millions -- this took the concept of heroism nowhere I'd ever imagined before.

Definitely the most memorable superhero movie *I've* ever seen, with indelible characters and performances to boot. For me, it was the best film I'd seen in 20 years.