3
$\begingroup$

Recently, I found out about somewhat famous Eliezer Yudkowsky and Machine Intelligence Research Institute he founded. Their philosophy and organisation seem interesting but I'm curious about their credibility. I'm pretty sure this is not a con and they seem to be producing a lot of articles. However, few of those are published and none in the journals mentioned here.

So, is MIRI doing genuine high-quality research?

Improve this question
$\endgroup$
1
  • $\begingroup$ A simple rule-of-thumb is to check where they publish their papers. For example, in total, how many papers have they published at top ML venues such as NeurIPS, ICML and ICLR? If the answer is none, they're probably not doing high-quality research. $\endgroup$
    – SpiderRico
    Jun 16, 2020 at 22:10

2 Answers 2

Reset to default
4
$\begingroup$

Whether or not it's produced good research is a question I can't answer. But I find it interesting and it seems reliable.

For instance, this paper on Parametric Bounded Löb's Theorem and Robust Cooperation of Bounded Agent is drawn from Cornell University's site. Likewise this recent paper on Logical Induction.

That said, they also seem to publish papers that have not been subject to peer review, which is an issue not specific to the site, and seems to more widespread phenomenon per easy access to digital files in general.

In relation to MIRI specifically, possibly this is a function of an an applied field advancing so rapidly that publication in journals is not seen as a priority, or even particularly relevant.

Improve this answer
$\endgroup$
3
  • $\begingroup$ Thank you for your answer. As far as I know anyone can publish on arXiv, as long as their paper satisfies formatting requirements. The fact that you find it interesting is a stronger evidence here :) $\endgroup$
    – Petr Naryshkin
    Jun 8, 2017 at 18:02
  • $\begingroup$ @ПетяНарышкин (This comes with the caveat that I am fairly disreputable;) I am noticing that with the the ability to access research papers online, vetting becomes a problem. I often come across papers have not been subject to peer review. Alignment for Advanced Machine Learning Systems, for example... $\endgroup$
    – DukeZhou
    Jun 8, 2017 at 18:06
  • $\begingroup$ that's one of my reasons to question MIRI. I don't know how it is done in CS and AI research but in math (my field) almost all good articles are published in one of the better journals. This makes it much easier to track the progress in topics you're curious about. $\endgroup$
    – Petr Naryshkin
    Jun 8, 2017 at 18:34
0
$\begingroup$

The utility of most of their papers is questionable. Evidence for that is that none of the proposed mechanisms (Tiling agents, Logical Induction, etc.) were to my knowledge applied to "real" aspiring proto-AGI systems ref. Sure some of their ideas may be good, time will be the judge.

Improve this answer
$\endgroup$

You must log in to answer this question.

Not the answer you're looking for? Browse other questions tagged .