@ks13735 @StefanFSchubert @Rx410 @Nonicoc No. Go take a look at the Wikipedia article for Finnish people.
English
Trevor Andersen
4.7K posts
@SkyNewsAust Chris Bowen said quite literally said, "buy as much fuel as you need", you goons
English
Anthony Albanese has burned through thousands of litres of jet fuel in just 10 days while his government suggests Australians limit their own consumption, triggering claims the Prime Minister is "out of touch".
skynews.com.au/australia-news…
English
@AntonyGreenElec @Qldaah Looks like we'll get a perverse outcome where the very likely Condorcet winner (Liberal) will lose. High primary Condorcet winners should not lose elections. The 3CP stage should be decided by the Condorcet criterion if all three candidates have a high 3CP share (say >25%).
English
The count in Hammond has been examining Airlie Keen's preferences. Not enough Keen flow to LIB to pass ALP so race looks like finishing as ONP v ALP. Election night count was ALP v LIB so new preference count needed. One Nation's Robert Roylance heavily favoured to win.
English
@LinkofSunshine Yeah the goal posts were moved to another stadium
English
@LocalBtnMasher Why does Crimson Desert, the biggest game, not simply eat the other games?
English
@ToryTheForg @morallawwithin @upsilon90794644 They'd be accused of racism and reverse polarise MAGA into veganism
English
@morallawwithin @upsilon90794644 I bet if they said "dogs aren't food" people would agree
English
Can someone please tell community notes about thick normative terms
PETA@peta
Chickens aren’t food.
English
@NZJennings @SashaGusevPosts I suspect some of Tao's outputs are beyond even my potential comprehension...
What, concretely, would qualify as a super intelligent mode?
English
@trevorandersen @SashaGusevPosts 2/2 What necessarily prevents qualitative transcendence is that the complete constraint resolution mode toolkit is already present in humans. Self-improvement calibrates it better. It can't add modes that don't exist.
English
Stumbled upon an interesting debate on AI super-intelligence from 2011. Yudkowsky makes three core claims/predictions, all of which are (to date) wrong:
1) That human intelligence is relatively simple and ASI can be achieved with a few small innovations; ...
English
English
@trevorandersen @kaiba_1991 @SashaGusevPosts Then you should specify the size of a napkin and how many characters can fit it (and also a size of those characters)
English
@NZJennings @SashaGusevPosts Terry Tao can probably produce outputs beyond my comprehension. Could a Tao-equivalent AGI qualitatively self-improve? I don't know, and I think maybe not, but I don't see what would *necessarily* prevent it?
English
@trevorandersen @SashaGusevPosts 2/2 Qualitatively superior cognition would produce outputs in principle beyond human comprehension. AlphaZero never did that. Superior access to patterns at speed, not superior cognition. The ceiling held.
English
@NZJennings @SashaGusevPosts Obviously not as general as intelligence, but did systems like AlphaZero not generate something like qualitatively superior "cognition"? I'm agnostic on RSI generating super intelligence and probably lean against it; I just don't see how we can yet rule it out.
English
@trevorandersen @SashaGusevPosts RSI during operation minimum required for genuine intelligence, = better equipped faster and better calibrated. What it can't do is generate qualitatively superior cognition, there's no mechanism for that, there's a ceiling. Which version of intelligence are you arguing for?
English
@duckduckcry @kaiba_1991 @SashaGusevPosts I also meant algorithms. I don't see that we can have any confidence yet that algorithms for generating true super intelligence will be napkin-expressible. The DNA that generates our brains probably has more Kolmogorov complexity than fits on a napkin.
English
@trevorandersen @kaiba_1991 @SashaGusevPosts I meant algorithms, of course.
And it’s true for any equations we have in the universe not counting singularities, collapse, and potentially fermion coupling
English
@duckduckcry @kaiba_1991 @SashaGusevPosts Most functions have Kolmogorov complexity exceeding the capacity of writing on a napkin
English
@kaiba_1991 @SashaGusevPosts 1) is meaningless. Any function can in principle be written on a napkin.
And besides it, your computer is not different in principle than a napkin and a pen, yk, so this a double-kill
English
@NZJennings @SashaGusevPosts It's a bit premature to conclude that RSI won't work
English
@SashaGusevPosts The predictions failed because the underlying model was wrong. Recursive self-improvement can't produce qualitatively superior cognition. There's a ceiling defined by a complete set of constraint resolution modes. Speed doesn't transcend that ceiling...
English
English
@trevorandersen @glow_guru @robertskmiles The way I use it is not, correct. I view this as a desirable feature rather than a bad thing. It enables trivial hard real-time guarantees to be made and for easy parallelization to occur, dramatically reducing V&V costs in safety-critical systems.
English
This is utter nonsense of course. The entire basis of the field of real-time computing is predicated on guarantees that computers will complete within a given deadline. We just bypass Turing’s proof by requiring a subset of code be used that has hard bounds on execution times.
vixhaℓ@TheVixhal
You cannot write a computer program that looks at any other program and correctly determines whether it will eventually finish or run forever. This isn't a hardware limitation. Alan Turing proved in 1936 that no such program can logically exist, by showing any attempt creates a paradox, which is essentially a formal version of “this sentence is false.” What's unsettling is the proof technique: he didn't find a hard case. He proved the question destroys itself when you try to answer it.
English
English
@glow_guru @robertskmiles The programs that can exist are relative to a particular kind of hardware. Change the hardware and you change the set of “all programs.”
English
The US government spent $25 million over a decade trying to prove your cell phone gives you cancer.
The study accidentally produced one of the strongest pieces of evidence for radiation hormesis ever recorded.
The NTP study was nominated by the FDA in 1999 specifically because they expected to find harm. They built 21 custom reverberation chambers in Switzerland. Exposed 1,679 mice and 859 rats to cell phone frequencies for 9 hours a day, every day, for 2 years. The whole operation was designed as the definitive “cell phones cause cancer” study.
The cancer results were mixed at best. Male rats got more heart schwannomas. Mice showed nothing significant. But the survival data was so unexpected that the researchers didn’t even know how to explain it in their own report.
Look at the survival curve. Every single radiation group outlived the control. The 2.5 W/kg group hit p=0.0020, the only statistically significant result in the entire longevity analysis. By day 700, the control group’s survival probability had dropped to ~0.65. The lowest dose group was still above 0.80.
That’s the hormesis signature. The smallest dose produced the largest benefit. The same pattern shows up in exercise, fasting, and cold exposure. A mild biological stressor activates repair mechanisms that wouldn’t otherwise turn on. Over 3,000 published papers have documented this across microbes, plants, insects, and mammals. The French Academy of Sciences formally accepted it in 2005. The US still builds its entire radiation safety framework on the opposite assumption: that all radiation, at any dose, causes proportional harm.
The FCC limit for cell phones is 1.6 W/kg. Your AirPods operate at a fraction of that. The dose that produced the strongest longevity signal in this study was 2.5 W/kg. Barely above the regulatory ceiling.
The entire regulatory framework for wireless device safety assumes a dose-response curve that this $25 million study failed to find.
Zane Koch@zanehkoch
for a while i've had a slight fear that the bluetooth from my airpods could be frying my brain this weekend i pulled the raw data from a $30m government study of 1,679 mice blasted with cell phone radiation and reanalyzed it what i found was...not what I expected? 🧵
English
@XiXiDu Two constant things are never "exponentially" related
English
Von Neumann's brain probably didn't have exponentially more neurons or consume exponentially more energy. Yet the difference between von Neumann and an average person is the ability to bring out the trash and contribute to the creation of superweapons.
Roko 🐉@RokoMijic
Theoretically speaking we should expect intelligence to have diminishing marginal returns. Every toy domain shows this. All machine learning models have strongly diminishing marginal returns on compute and model size. There are no exceptions. Runaway recursive self improvement has no empirical or theoretical support.
English
@trevorandersen @drorpoleg @DouthatNYT If I said somebody could be a king of the cannibals it would be at least 5-10% extremely dry insult.
English
What are we even doing here?
x.com/sama/status/20…
Sam Altman@sama
Tonight, we reached an agreement with the Department of War to deploy our models in their classified network. In all of our interactions, the DoW displayed a deep respect for safety and a desire to partner to achieve the best possible outcome. AI safety and wide distribution of benefits are the core of our mission. Two of our most important safety principles are prohibitions on domestic mass surveillance and human responsibility for the use of force, including for autonomous weapon systems. The DoW agrees with these principles, reflects them in law and policy, and we put them into our agreement. We also will build technical safeguards to ensure our models behave as they should, which the DoW also wanted. We will deploy FDEs to help with our models and to ensure their safety, we will deploy on cloud networks only. We are asking the DoW to offer these same terms to all AI companies, which in our opinion we think everyone should be willing to accept. We have expressed our strong desire to see things de-escalate away from legal and governmental actions and towards reasonable agreements. We remain committed to serve all of humanity as best we can. The world is a complicated, messy, and sometimes dangerous place.
English
@joquji5n324il @drorpoleg @DouthatNYT Very obviously a compliment given their business relationships
English
@drorpoleg @DouthatNYT Did he mean it as a compliment or an insult? 🤔
English
@LoganView64206 @Plinz But why would they think to save resources in this particular case?
English
@Plinz Could this just be a resource minimization issue? IE, most models just jump to the first answer (the riddle) to save resources, rather than fully evaluating the question for possible additional complexity?
English
