MrThink

Karma: 297

MrThink May 28, 2022, 9:57 AM
1 point
in reply to: MrThink’s comment on: [$20K in Prizes] AI Safety Arguments Competition
For researchers (mainly)
Artificial intelligence isn’t limited in the same ways the human brain is.
Firstly, it isn’t limited to only run on a single set of hardware, it can be duplicated and speeded up to be thousands of times faster than humans, and work on multiple tasks in parallel, assuming powerful enough processors are available.
Further, AI isn’t limited to our intelligence, but can be altered and improved with more data, longer training time and smarter training methods. While the human brain today is superior to AI’s on tasks requiring deep thinking and general intelligence, there is no law preventing AI’s from one day surpassing us.
If artificial intelligence were to surpass human intelligence, it would likely become powerful enough to create an utopia lasting a long long time, or spell the end of humanity.
Thus, doing AI safety research before such an event becomes vital in order to increase the odds of a good outcome.

MrThink May 27, 2022, 2:01 PM
1 point
in reply to: MrThink’s comment on: [$20K in Prizes] AI Safety Arguments Competition
What if AI safety could put you on the forefront of sustainable business?
The revolution in AI has been profound, it definitely surprised me, even though I was sitting right there.
-Sergey Brin, Founder of Google
Annual investments in AI increased eightfold from 2015 to 2021, reaching 93 billion usd.
This massive growth is making people ever more dependent on AI, and with that potential risks increases.
Prioritizing AI safety is therefore becoming increasingly important in order to operate a sustainable business, with the benefits of lower risks and improved public perception.

MrThink May 27, 2022, 1:45 PM
1 point
in reply to: MrThink’s comment on: [$20K in Prizes] AI Safety Arguments Competition
For researchers
What if you could make a massive impact in a quickly growing field of research?
As artificial intelligence continues to advance, the potential risks increase as well.
In the words of Stephen Hawking “The development of full artificial intelligence could spell the end of the human race.”
AI safety is a field of research with the purpose to prevent AI from harming humanity, and due to the risks current and future AI is posing, it is a field in which researchers can have a massive impact.
In the three years from 2016 to 2019, AI research has grown from representing 1.8% to 3.8% of all research papers published worldwide.
With the rapid growth of general AI research, we should expect both the importance as well as the funding for AI safety to increase as well.
Click here, to learn how you can start your journey to contribute to AI safety research.
Notes
Depending on context, this text can easily be shortened by removing the third to last and second to last paragraphs.
I used this graph to get the growth of AI publications:

MrThink May 3, 2022, 3:33 PM
6 points
in reply to: MrThink’s comment on: What would a 10T Chinchilla cost?
I believe that given a few years, a company wanting to make a 10 trillion parameter GPT-3 could probably do it for less than these estimates, since 71-358 million usd isn’t that much in compute, and for that money extra specialized hardware produced in bulk could be used to bring costs down.

MrThink May 3, 2022, 3:31 PM
25 points
on: What would a 10T Chinchilla cost?
Chinchilla is 70 Billion parameters. 10 trillion / 70 billion ≈ 143
According to the paper released with the information about Chinchilla (Training Compute-Optimal Large Language Models” by Hoffmann et al.), they claim ““For every doubling of model size the number of training tokens should also be doubled.”
So if this would be followed, 143 times more data would also be needed, resulting in a 143*143= 20449 increase of compute needed.
Chinchilla probably costed around 1-5 million usd in compute power to train, the cost for a 10 trillion parameter version would cost around 20.5 billion to 122.5 billion usd.
However, this is not a viable alternative, because there is not enough text data available.
A more realistic scenario is to perhaps double the data Chinchilla was trained on (which might not even be easily done), and then 143x the size, so a cost of about 286 million to 1.43 billion usd.
Keep in mind, that GPT-3 was trained on about a quarter of the data that Chinchilla was trained on. So a 10 trillion parameter GPT-3 model might cost around 71.5 million to 358 million usd.

MrThink May 2, 2022, 12:06 PM
2 points
in reply to: MrThink’s comment on: [$20K in Prizes] AI Safety Arguments Competition
The main source of skepticism of AI safety research, is that it’s unknown how advanced current AIs are and how fast it is improving. The most impressive reasoning task I have seen an AI model do is this one, done by PaLM by Google:

The model was prompted with this text:
Michael is at that really famous museum in France looking at its most famous painting. However, the artist who made this painting just makes Michael think of his favorite cartoon character from his childhood. What was the country of origin of the thing that the cartoon character usually holds in his hand?
And the model answered this:
The most famous painting in the Louvre is the Mona Lisa. The artist who made the Mona Lisa is Leonardo da Vinci. Leonardo da Vinci is also the name of the main character in the cartoon Teenage Mutant Ninja Turtles. Leonardo da Vinci is from Italy. The thing that Leonardo da Vinci usually holds in his hand is a katana. The country of origin of the katana is Japan. The answer is “Japan”.
This example can be used in a range of situations.

MrThink May 1, 2022, 9:35 PM
1 point
in reply to: MrThink’s comment on: [$20K in Prizes] AI Safety Arguments Competition
For tech executives
Could working on AI safety put you on the forefront of sustainable development?
As AI is becoming increasingly advanced and relied upon, the recognition of the importance of AI safety is increasing. If this trend continues AI safety will likely become a core part of sustainability, and the businesses that prioritizes AI safety early, will have a more sustainable business as well as a positive impact and improved public perception.

MrThink May 1, 2022, 9:08 PM
1 point
in reply to: MrThink’s comment on: [$20K in Prizes] AI Safety Arguments Competition
As AI technology is heading towards becoming sophisticated enough to potentially end civilizations, the impact and recognition of those working on AI risks will increase.

MrThink May 1, 2022, 8:34 PM
1 point
in reply to: MrThink’s comment on: [$20K in Prizes] AI Safety Arguments Competition
General paragraph for non-technical people
What if you could increase your impact by staying ahead of the AI trend?
In a hundred years computers went from being punch card machines, to small tablets in every pocket. In sixty years computers went from displaying only text, to providing an entire virtual reality. In the past three years, AI has become able to write engaging stories and generate photorealistic images.
If this trend continues, AI is set to cause massive change. If this change is positive or negative depends on what is done today. Therefore actions taken today have the potential of massive impact tomorrow.
Why this paragraph?
Most descriptions focus on how terrible AI can be, and fails to convey what the person reading has to gain personally by taking action. Having impact is something most people desire.
Depending on context, the paragraph can be tweaked to include what action the reader should take in order to have massive impact.

MrThink May 1, 2022, 8:34 PM
2 points
in reply to: MrThink’s comment on: [$20K in Prizes] AI Safety Arguments Competition
AI is like fire, we can use it to enter a new era of civilization, or to burn everything down.

MrThink May 1, 2022, 8:34 PM
1 point
in reply to: MrThink’s comment on: [$20K in Prizes] AI Safety Arguments Competition
If AI is set to cause massive change, acting now will have a massive impact on the future.

MrThink May 1, 2022, 8:33 PM
1 point
on: [$20K in Prizes] AI Safety Arguments Competition
I will do as Yitz post my submissions as individual replies to this comment. Please let me know if there’s any issues with that.

MrThink Apr 18, 2022, 6:02 PM
5 points
on: The Efficient LessWrong Hypothesis—Stock Investing Competition
Winner announcement!
The competition has now been finalized.
No comment with a security suggestion reached five karma, but to encourage participation in future contests, I’m lowering the limit to three karma. That means that jimv and cat from /dev/null are winners!
Also, lechmazur also wins 30 usd for his comment with 11 upvotes.
Winners will be contacted shortly.
Soon part two of the competition will start. Participants will be able to probability distributions on the securities from this contests. Rewards will be given to participants in relation to how much their prediction outperforms the market.

MrThink Apr 17, 2022, 9:45 AM
2 points
in reply to: MrThink’s comment on: The Efficient LessWrong Hypothesis—Stock Investing Competition
5. The current rating system isn’t particularly good, most people (including myself) gives 5 stars all the time unless something was wrong, so almost all sellers have like 5 or 4.9 stars. However it is possible that Fiverr can improve the rating system, to make it a much better predictor of how well a seller will be at performing a certain task. It could be a huge value add, since oftentimes the price and quality for consultancy services can vary hugely. Just look at video on youtube where they pay people on Fiverr to do stuff like edit videos, play instruments etc. The correlation between price and quality is quite low.
6. I have used Fiverr a lot, and in my experience the results has generally been better than other platforms.

MrThink Apr 13, 2022, 6:59 AM
2 points
on: The Efficient LessWrong Hypothesis—Stock Investing Competition
Suggestion: Fiverr International Ltd.
Reasons:
1. Jobs are becoming more and more specialized. This is a trend that has been lasting for thousands of years, and likely won’t stop anytime soon. Fiverr is a platform for buying and selling gigs, allowing for extreme specialization.
  
  I think the market values Fiverr primarily on current numbers on growth. If we predict that the market will eventually blow up, this would likely lead to a growth acceleration of fiverr that the market might not have anticipated.
2. Small market cap (2.3 Billion USD) - Smaller businesses are less analysed by professionals, and therefore should in theory be easier to make good investments in.
  
  Also, a small market cap means massive growth potential. If buying gigs become a massive industry, and big companies start creating systems for buying gigs. It seems possible for Fiverr to 10x-100x.
  
  Many other platforms for buying and selling has market caps, ex: Airbnb ($103 billion), Uber (64$ billion)
3. Good numbers. Fiverr has a price-to-sales (PS) ratio of 8.1 (source). This is significantly better than Airbnb at 17.08 (source). For comparison, the S&P 500 has a PS ratio of 2.91, so if Fiverr were to just over double it’s revenue, it would be on par with the S&P.
  
  Fiverr is a growth company, so I don’t understand why the market things it should have such a low PS ratio, but I’m sure it has it’s reasons. But the low PS ratio is still a good sign, since it proves that there seems to be fundamental value to invest in, and not just a good idea.
4. Profitable? It depends on how profit is measured (so no not really profitable), but at least it doesn’t seem like Fiverr is pumping out money like crazy, which suggests that the risk of going bankrupt isn’t super high.
Personal thoughts: While I do believe the EHM to be true. I expect cases like Bitcoin and potentially Fiverr, to be opportunities where we can predict massive growth, even if the odds are only a few percent, to be one of our best bets at beating the market. I might suggest diversifying the risk a bit by investing in some competitors too, like Upworks, but if I would only bet on one stock it would be Fiverr.

MrThink Apr 12, 2022, 9:32 PM
2 points
in reply to: Rossin’s comment on: The Efficient LessWrong Hypothesis—Stock Investing Competition
Alibaba does looks cheap, with a PE ratio of 15, and probably likely to grow faster than the S&P 500.
However, I am not sure why the market would be wrong about the risks of delisting, regulation and instability in China.
I think one not obious advantage with Alibaba is that it’s one of the strongest AI companies in China, and China is investing a lot in AI, and might (not sure how likely) become the biggest player in AI globally.

MrThink Apr 12, 2022, 7:09 AM
1 point
in reply to: jimv’s comment on: The Efficient LessWrong Hypothesis—Stock Investing Competition
Those seem like reasonable arguments.

It might be worth to point out, that Tesla has had way smaller budgets for R&D than other manufacturers like Ford and GM. So if Tesla holds an advantage, it implies that simply investing more money won’t let other manufacturers catch up.

MrThink Apr 12, 2022, 7:04 AM
1 point
in reply to: Lech Mazur’s comment on: The Efficient LessWrong Hypothesis—Stock Investing Competition
Good point.

I think for the second competition where development will be predicted fees should be included to more accurately reflect the reality.

In this competition I think disregarding fees might still be viable, since it’s mainly people’s reasoning that is interesting.

MrThink Apr 12, 2022, 7:00 AM
1 point
in reply to: trevor’s comment on: The Efficient LessWrong Hypothesis—Stock Investing Competition
Is the main argument then that higher uncertainty, generally leads to undervaluations of assets?

In general more high risk investments should be accompanied by higher expected returns, which might be the argument here?

MrThink Apr 12, 2022, 6:57 AM
1 point
in reply to: Zach Stein-Perlman’s comment on: The Efficient LessWrong Hypothesis—Stock Investing Competition
Google has accounted for several of the biggest AI breakthroughs in the past 10 years, perhaps most notably, the creation of the transformer architecture. Further they invest a lot in AI, which among other things becomes apparent by them creating TPUs (their own hardware for AI).

Thus I believe Google to be the most likely candidate for achieving future breakthroughs in AI. It is not yet clear to me however if they will manage to capitalize on it well, since in they usually share all their breakthroughs openly.

Further, Google has a PE of 23.14, well below S&P 500’s PE of 24.91. Meaning that of Google has the equal or more growth potential than the S&P 500 in general, it should outperform.