DennisP

67 ms·

961.
by DennisP 3y ago
This looks like a big advance in alignment research. A big problem has been that LLMs were just a giant set of inscrutable numbers, and we had no idea what was going on inside. But if this technique scales up, then Anthropic has fixed that.
962.
by DennisP 3y ago
Yes, but we have no idea what the probabilities are. We have no basis for saying "it is probable that Mars had life." If we found native life on Europa or something, that would change things completely. Independent life arising at
963.
by DennisP 3y ago
We can't know that life is guaranteed unless we find another example of life arising independently. We don't know how many random chemical reactions are required for life to be likely. We probably agree that a small puddle isn&#x2
964.
by DennisP 3y ago
But we also don't know how life started. There's a big gap between the simple organic molecules that pop up everywhere, and metabolizing reproducing life, and we have no idea how many improbable events had to occur to bridge that
965.
by DennisP 3y ago
I'm a big fan of his book The Wreck of the River of Stars , and think a lot of people here would like it. The River of Stars is an interplanetary craft, originally driven by solar sail, with snazzy new fusion drives recently added.
966.
by DennisP 3y ago
D-T: Where you get all those energized neutrons is easy: from D-T fusion. You need a neutron multiplier like lead or beryllium, then it looks like you can average about 1.15 tritium atoms per D-T reaction. Helion: to use their tritium they
967.
by DennisP 3y ago
D-D requires a higher triple product than D-T fusion and produces less energy, so most companies go with D-T, and don't attempt to use D-D to produce the tritium. But D-D takes a lower triple product than D-He3, so if you can get n
968.
by DennisP 3y ago
Fine. If you don't have evidence either way, might as well go with option two.
969.
by DennisP 3y ago
If management can tell which way gets more work done, then they don't have to rely on workers being honest about it. If they can't tell the difference, seems to me they should let people have nicer lives.
970.
by DennisP 3y ago
In sunlight you also get lots of red and infrared, and there's quite a bit of research on the benefits of that. To replicate sunlight you'd also need bulbs for those frequencies.
971.
by DennisP 3y ago
It's just temperature, density, and confinement time. For each fuel, there's a minimum threshold temperature and a minimum triple product of those three parameters to get net power. The triple product is lower for D-D than it is f
972.
by DennisP 3y ago
> Base it on say, Karl Popper and Media Literacy theory How would this work?
973.
by DennisP 3y ago
I haven't noticed that but it's been quite a while since I cooked unprocessed raw pork. I wonder if the intense ammonia smell that builds up in pig factories had anything to do with the smell you experienced.
974.
by DennisP 3y ago
Nobody is prohibited from using the name "parmesan," which is a relatively recent term. That's why you can buy powdered "parmesan" from your grocery store that's up to 8% cellulose and made in the US. People ar
975.
by DennisP 3y ago
You don't need to breed He3 unless you can get net power from D-He3. If you can do that, then you can probably also get net power from D-D, which doesn't require as high a triple product. The fusion company Helion is working on ex
976.
by DennisP 3y ago
I'm just saying $3 million doesn't do much to the price. If someone suddenly cashes out several thousand times that much, then sure, price will crash. I've personally watched sell pressure on Coinbase in the multi-million ran
977.
by DennisP 3y ago
Nope, daily Bitcoin volume on Coinbase alone is over $200 million. Nobody cares if you sell 1% of that, at worst the price dips momentarily.
978.
by DennisP 3y ago
That's not an actual problem. To breed tritium, bombard lithium with the neutrons from D-T fusion. To breed helium-3, fuse deuterium.
979.
by DennisP 3y ago
The difference from cold fusion is that NASA is blasting the lattice with gamma rays, which is what gives deuterons the energy to fuse. Whether there's a chance of getting net power that way, I have no idea, but at least it's not
980.
by DennisP 3y ago
> Now we have this... ODIN—a way to outsource your intellectual note-taking procrastination to AI. I'm not sure but it sounds like you're scoffing at that. To me it sounds like the solution to people spending too much time ta
981.
by DennisP 3y ago
Perfect Day has engineered yeast to make milk proteins. They're selling ice cream made of actual milk, produced from yeast.
982.
by DennisP 3y ago
The fruit hung lower back then but now we have taller ladders. We've learned a lot since the 1960s. That blog entry is interesting but it's based on an article Malcom Gladwell wrote about DeVita's book; the author says he had
983.
by DennisP 3y ago
Everybody's terrified of the snake oil days, and almost nobody considers the way things were several decades ago, when cancer researchers were allowed to try things on patients without government approval, and we made much faster progr
984.
by DennisP 3y ago
It's sad what happened at Huy Fong. For years they got fresh peppers from a nearby farm. Then they blew up that relationship, perhaps because they decided they could get cheaper peppers from Mexico. Now the original farm still happens
985.
by DennisP 3y ago
It sometimes startles me how disconnected some people are from nature. I can't imagine all our pine trees dying off, it seems utterly horrific. It would be devastating to biodiversity and I would deeply miss the trees themselves. But t
986.
by DennisP 3y ago
Yep it's great for work emails. Incoming too, since they can summarize a long email into bullet points. The future is people typing bullet points, expanding into polished prose for transmission, and compressing down to bullet points on
987.
by DennisP 3y ago
They're only net carbon emitters if they're powered by fossil fuels. The Climeworks plant is powered by geothermal. Any renewable or nuclear power source would be fine. It's expensive but it's scalable. There is a vast s
988.
by DennisP 3y ago
Maybe it's not done yet.
989.
by DennisP 3y ago
In general, I think we should reserve the term "greenwashing" for fake projects. Lots of carbon offsets, for example, actually do nothing at all. If you're using those projects to offset your personal emissions, you're w
990.
by DennisP 3y ago
Not entirely: > in the UK, the least wealthy half of the population accounts for less than 20% of final demand, less than the top 5% consumes.

More ›