Jump to content


Senior Members
  • Content Count

  • Joined

  • Last visited

  • Days Won


Prometheus last won the day on February 25

Prometheus had the most liked content!

Community Reputation

561 Glorious Leader


About Prometheus

  • Rank

Profile Information

  • Interests
    Building statistical models for Raman spectroscopy.

Recent Profile Visitors

21797 profile views
  1. I find a useful way of thinking about evolution is as an optimisation problem - species are 'tweaking' various parameters in order to improve fitness, which is a kind of objective function. However, that objective function is itself subject to change because it is dependent upon the environment and ecosystem which are themselves constantly changing.
  2. No, you just take into account the numerous wet markets in Wuhan which bring together humans and various animals. Then ask, what is the simpler explanation (since you invoked occam's razor) - that a virus mutated naturally or it was lab created. A natural evolution sounds simpler to me.
  3. Pandemics are replete throughout human history - occam's razor in this context would be that this is another natural pandemic.
  4. I think they could replace current peer review. A community much like that at Cross-Validated could provide the peer-review, with a public back and forth between commentors and the authors discussing various points. It would give far more transparency to the process. To prevent spamming could use some machine learning to filter out absolute junk and also have submissions cost a little bit of reputation (rep being earned by getting actually published, usefully reviewing others work, maybe accounts linked to a uni get to start with a few free rep points). Once it's been through the preprint peer
  5. If you're not proposing a change in architecture to deal with this, i think it's slightly off topic, but it's an interesting tangent. Say you have a language model and you start talking about a doctor and ask the model to complete the paragraph. A fair model might be one that goes on to use male and female pronouns in equal amounts. But how to achieve that end? I don't think filtering the dataset is practical. GPT-3 was trained on words scraped from the internet, something like 300 billion 'tokens' (token ~= word). There's no way even a team of humans could curate that. You could try
  6. Interesting point. But those biases exist because the training data itself was biased - misclassifying people of minority groups because the data was trained on the majority group. Essentially it was shown unrepresentative training data given tasks it had not be trained on. Additional architecture shouldn't be required to remedy this bias, just appropriate data curation, and thorough testing before deployment. I wasn't aware i was doing so. Does that mean you advocate for more flat architectures?
  7. True enough, but i thought it a good case study - better the Iron Mike anyway. This list has about 40 transgendered sports people. Whether that's enough to ask questions is a judgement call, but i don't think it unreasonable of professional sporting bodies to pose the question to the medical community.
  8. Ronda Rousey is generally considered the best female MMA fighter of all time. Her record was 12 wins from 14 fights. By that metric i'd consider Fox's 5 wins out of 6 at least moderately successful.
  9. I agree, but the counter argument is that so far the biggest improvements have simply been bigger models, not more structured models. There was a more subtle point at the end of the article suggesting that we shouldn't be trying to inject human knowledge directly into AI because the mind is complex beyond our understanding and if we don't know how knowledge is codified in the brain, we don't have a basis to codify it in AI. I don't understand what you're trying to say.
  10. I gave this example early on in the thread; it raised questions in MMA. I agree that Curious laymen was unfairly set upon in this thread.
  11. I learnt classical stats so i'm in no position to defend Bayesian approaches, but i'm not sure we need to dump Bayesian techniques. They are both tools and some jobs may be better suited to one or the other. I know there are a lot of physicists on this site so i wonder what they would make of a Bayesian approach to something like quantum theory. Also, something i've never understood: both approaches are consistent with the Kolmogorov axioms, so in what sense are they really different at a fundamental level? I'm surprised this topic hasn't reared its head on this forum before.
  12. The biggest lesson that can be read from 70 years of AI research is that general methods that leverage computation are ultimately the most effective, and by a large margin. According to Deepmind's Richard Sutton the bitterness is that injecting human knowledge into an AI systems results only in short term gains and that general statistical methods that leverage computational power perform much better in the the mid to long term. GPT-3 has 175 billion parameters and doesn't seem to be approaching the limit of what a simply bigger model can do. Some speculate GPT-4 will have trillions of p
  13. So we have all these ones and zeros in a superposition of states, but for the final output we want it to collapse to a single state: the right answer. How is this collapse controlled to that end? Or is this not how it works.
  14. As far as i can tell from the BMJ editorials related to those studies, the concern in sports medicine seems to focus mostly on the nuances of competition rules. For instance, there are calls to make the period of hormone therapy longer before allowing transwomen to compete with ciswomen. There have also safety concerns raised in the MMA community, though i can't find anything related to that with my glance at the academic literature. Haven't time for a deeper dive so i'll have to leave it there.
  • Create New...

Important Information

We have placed cookies on your device to help make this website better. You can adjust your cookie settings, otherwise we'll assume you're okay to continue.