how would LLMs help with uncovering "bad studies"? I don't think LLMs are yet sophisticated enough to figure out potential plagiarism or Adobe Photoshop copy-and-paste (as mentioned in the article above).
550 karma · joined July 14, 2021
Is there a reason TECH companies want old very-likely-tech-illiterate politicians to join their board? Is the only reason to be more cozy with washington?
Fun fact : one method of drug transfer (mentioned in popular media) from Pakistan (bordering Punjab) to Punjab is discus throw [2]. Here, a local athlete takes a disk of drugs and throws it over the border which is then intercepted by dealers in Punjab. A great movie about the drug trade in Punjab is "Udta Punjab" (literally translated as flying-Punjab or high-Punjab).
I can only imagine how many drones they missed. Drones usually fly at an altitude that our eye can't see. The drone operators must be smart enough to fly during cloudy days and during night time making binoculars useless as well. My best guess would be ground intelligence by police.
I am a bit surprised that drug trade in a state in India is using drones. The cost of drones (even commercial ones cost ~$500) is a pretty steep one when converted to Indian currency (~40,000 INR). The article[1] states that the police shot down >100 drones = INR 40,00,000. There must be a lot of money in local drugs in Punjab for the operators to be willing to sacrifice 100+ drones.
[1] https://www.hindustantimes.com/cities/chandigarh-news/107-dr...
New technology used for war such as lasers or AI will only make war even more asymmetric.
So essentially, big-countries can't attack other big-countries because of nukes and MAD (mutually-assured-destruction). Big-countries can't attack small-countries because of asymmetric war. Small-countries can't attack big-countries for obvious reasons. So, the only war left is small-countries vs small-countries.
i'm curious how the author felt knowing that his paper would be so relevant only a few months after publication.
also, is it possible to see distribution of users based on geography? would be very interesting
And then if we put all this data into an anomaly detection algorithm, will the Earth stand out? Based on the data that we capture, are we actually special?
The reason for anomaly detection (unsupervised) alg is that it doesn't have any of our subjective biases.
The answer can give evidence for whether seeing signs of interesting planets is actually sign of life or not.
Thoughts? I don't think Earth will stand out based on the information that will reach tens of light years away.
thoughts from anyone proficient in statistics?
They are being intentionally vague here. Define "most powerful". And what do they mean by "share". Do we need approval or just acknowledgement?
This line is a slippery slope for requiring approval for any AI model which effectively kills start-ups, who cannot afford extensive safety precautions
Do people stop talking to him in fear of saying something bad? Do his old friends stop contact because of what he might have told FBI? Does he stop getting invited to parties?
Or will all of this blow over in a couple weeks and it's business as usual? My thinking is that his social life will change quite a bit (even accounting for him being a billionaire) and he is done with politics. I don't think any politician will want to involve themselves with Thiel now.
People aren't money but money can be converted to saved lives.
For the crime that SBF is alleged to have committed, we can reasonably expect 20 yrs (he will be out in 10-12). But if he had committed a murder, he would be in prison for life (no parole).
Mathematically speaking, this implies that a life is worth more than fraud worth 8 billion dollars. We can make a very reasonable assumption that the govt can save 1 additional life using a million dollars (giving necessary treatment to drug addicts / suicidal people / etc). Out of the 8 billion dollars from SBF, government would have got at least 50 million dollars from taxes (this is a lower bound). Out of that 50 million, maybe the govt puts 10-20 million back for public good. So the govt could have saved at least 10-20 people from the money that SBF is alleged to have defrauded.
Based on the above, can we say that SBF indirectly killed 10-20 people? What would be the punishment for someone who kills 10-20 people? Surely more than 20 years.
This balance between the scale of crime and the severity of punishment can be very interesting. Should someone defrauding 2 billion dollars suffer twice the punishment of someone defrauding 1 billion dollars? What about 2 murders to 1 murder? What about numbers on scale that can change an economy (what SBF did)?
More fundamentally, is our legal system scalable?
My guess would be that engineers first go to documentation, don't understand it, go to low-quality-content-farms which answer their questions in natural language. It's low quality but it's enough for novice use cases such as python endsWith.
And this leads to a big reduction in Google's ranking of python docs.
TLDR : most novice programmers don't/can't read docs.
I have always found concepts such as "commoditize the complement" very interesting. For some reason, they seem like pseudo explanations of behaviours we see from companies. I am not entirely convinced that this was Meta's reasoning for releasing Llama. Like, I think Zuck is not thinking along these lines while open srcing LLMs.
How is Llama a complement of Meta's products? Meta wants more social interaction. It's a big stretch to assume LLMs will lead to more shareable social content (unless meta wants AI bots, in which case they don't need LLMs; they can just relax their moderation policies).
I honestly still don't have any good explanations for releasing Llama. Even the explanation of Meta gaining value from public improvements to Llama seems too vague and a stretch. Meta has enough engineers to make the most useful improvements themselves. The cost and effort of open srcing Llama is >> the value Meta gains from public improvements of Llama.
So, yeah this is still an open ques for me...
I am not entirely convinced that this was Meta's reasoning for releasing Llama. Like, I think Zuck is not thinking along these lines while open srcing LLMs.
How is Llama a complement of Meta's products? Meta wants more social interaction. It's a big stretch to assume LLMs will lead to more shareable social content (unless meta wants AI bots, in which case they don't need LLMs; they can just relax their moderation policies).
I honestly still don't have any good explanations for releasing Llama. Even the explanation of Meta gaining value from public improvements to Llama seems too vague and a stretch. Meta has enough engineers to make the most useful improvements themselves. The cost and effort of open srcing Llama is >> the value Meta gains from public improvements of Llama.
So, yeah this is still an open ques for me...
The requirements seem very subjective (define extraordinary). And so I wanted to know what are some real requirements/conditions the committee uses in real life for people who have already got O-1s.
I am a ML researcher (with a couple papers) in Canada (intl student from India), in senior year of undergrad. What are my chances of O-1? It's supposed to be for extraordinary people which, by definition, can be a very high bar to clear.
So I was wondering if there is any point working towards it or just go with L1B or something?
Any information would be appreciated. Thanks
I want to measure my pupil distance while browsing social media. The reasoning is my reading in "Thinking Fast and Slow" that pupils dilate when we see something we find interesting / something we like / when we are thinking and vice-versa. I want to put it to the test and it seems like finally the consumer tech is close to making it possible.
Does quest support pupil tracking? I did some cursory research but couldn't find any reference for it. Industry pupil tracking headsets are way too expensive; My last hope is that someone will jailbreak vision pro...
I want to measure my pupil distance while browsing social media. The reasoning is my reading in "Thinking Fast and Slow" that pupils dilate when we see something we find interesting / something we like / when we are thinking and vice-versa. I want to put it to the test and it seems like finally the consumer tech is close to making it possible.
Does quest support pupil tracking? I did some cursory research but couldn't find any reference for it. Industry pupil tracking headsets are way too expensive; My last hope is that someone will jailbreak vision pro...
I have never seen a service/industry move from ads to subscription. People will watch as many ads as you throw at them to avoid paying.
For example, even though youtube keeps increasing ads to an annoying level, vast majority of public is willing to grovel through that rather than pay for it. What evidence do you have that makes you think that that will change?
Please don't get me wrong. I admire someone creating a new and different search engine but if you keep growing in size, I believe you will face the same pressures that google/bing/etc faced and go the same route they did. "The king is dead, long live the king" scenario.
PS : remember larry and sergey warned about advertising incentives in their research paper as well but once you see the 9-10-11-12 digit figures on a piece of paper, idealistic morality usually goes out the window.
Ofc this is just my "educated guess at best, speculation at worst". I have a lot less information than you in this field.