timy2shoes

What you call bugs I call features. See e.g. https://news.ycombinator.com/item?id=25443212

Yeah, they had one of the best remote talks I've ever seen at last year's NeurIPS: https://nips.cc/virtual/2020/public/invited_16166.html

It depends. Say in the case where you quit your current job, or move, or anything where you've spent or lost money based on a promise of a job, then you might be able to recover money on the basis on Promissory estoppel: https://www.law.cornell.edu/wex/promissory_estoppel

The problem with defining clear goals for Quantum AI is that if you try to measure them then they will change.

The start of this line of thought was Robert Bork in 1978: https://en.wikipedia.org/wiki/The_Antitrust_Paradox

Yes, but that would require knowing how to rescue the phenotype, in this case healthy neurons and connections. Alzheimer's is not like Parkinson's, which you can (most likely, see the recent work from BlueRock) treat by replacing dopaminergic neurons or rescuing the dopamine production.

With a limited NIH budget this will necessarily divert funds from other endeavors. In addition, I think such programs entrench the large lab bias prevalent in academia right now.

I think the overall goal assumes that genetics causes Alzheimer's, which is (in my opinion) a questionable assumption. That's not to say that the production of model systems for Alzheimer's will be useful, but will it be worth the tens or hundreds of millions of dollars that this will cost? ¯\_(ツ)_/¯

Oversampling, then applying a train-test split? Jesus, that's like machine learning 101. But then again, I see a lot of questionable practices in the application of ML in biology.

Won't non-intact mRNA without a stop codon get degraded by Non-stop decay?

> Imaging based diagnosis could read presence or absence of particular gene mutations from the images so that the genes can be silenced by the drugs.

>Imaging based diagnosis could also figure out whether a particular cancer precursor is going to develop into invasive cancer and do it better than the experts we have now (otherwise we wouldn't use the AI).

Where is the evidence for these claims, other than a VC hype sheet? Like real clinical trials. These claims also show a fundamental misunderstanding of what this data can tell us. Imaging data doesn't give you tumor genetic profiles. It can give you tumor phenotype, which is associated with specific mutations. To get the true genetic profile you need to do deep sequencing at tens of thousands of dollars per tumor, and even then you have the problem of tumor heterogeneity, which lets the cancer evade the treatment.

A major concern I have working in this space is that we're selling people on grand promises of far off possibilities rather than what we can actually deliver right now.

Cas13 doesn't target DNA, rather targets RNA. It cuts the RNA, which then degraded. See https://www.sciencedirect.com/science/article/pii/S009286742...

Exactly. It's difficult to make causal claims from observational data, and made worse by the constant flood of causal claims coming out of microbiome labs. At best there's anecdotal evidence, at worst researchers are peddling snake-oil.

And to clarify, I have worked on microbiome research. The whole field is a mess.

Converting text to numbers via individual characters destroys any semantic meaning to the input. Which is why no advanced NLP tools will do such a thing. Basically, what they describe takes whatever meaning is in the title and throws it in a shredder.

Title A: Show HN: New grep tool written in Python

Title B: Show HN: New grep tool written in Garbage (1% better)

I guess Python is 1% worse than garbage. Lol

Yeah, this is a classic case GIGO.

> Title A: kdjfs;dfisd;diofuosd

> Title B: sdlkn nmkd kljdsa;'[ (10% better)

> The winner is title B: sdlkn nmkd kljdsa;'[

There's so much more missing that makes me think this is an engineer writing about how biological research should be in theory, without the understanding and experience the complicated nature of biological research. Well, suppose we want to give "every drug, at every dosage, every regiment, and in every combination to a mouse". Well, that's a combinatorially large search space and one reason drug companies focus on high throughput screening of drugs before applying them to mice. Then even for the promising ones you have other effects that will skew your results, like handler effects (e.g. https://www.nature.com/news/male-researchers-stress-out-rode...). If we want to apply the drugs to that many mice, are we only gonna use female mice because males tend to fight and kill each other when placed in the same cage? Then what do we do about the sex skew? Then there's the fact that 96 million years separate humans from mice, which is relatively short in a genetic perspective but not that much from an epigenetic perspective. How well will results translate?

This is just off the top of my head, and I'm not a biologist. I'm sure actual experienced biological researchers can come up with a whole bunch of other issues. Because more often than not biology is whole lot messier than we can imagine.

We can't have large investors losing money on any of their (inherently risky) investments. How will the market work if investors can lose money?

Yes, and it's difficult to get information from charter schools because they are not subject to the same reporting that public schools are. This makes it easy for people take advantage of their opaqueness.

These are a few examples from Los Angeles:

https://www.washingtonpost.com/education/2019/06/17/how-does...

https://patch.com/california/studiocity/la-charter-schools-f...

https://www.scpr.org/blogs/education/2013/04/05/13188/los-an...

You can open it in incognito mode to avoid the paywall.

If y = 2^x, then log y != 2^log x because of Jensen's inequality. You can take log of both sides to keep the equality, but applying the transformation inside of 2^() results in the equality turning into a strict inequality.

Exactly. Biases in previous data will be amplified with predictive policing algorithms. See, for example, https://rss.onlinelibrary.wiley.com/doi/full/10.1111/j.1740-...

Seems pretty straight forward. 162,000 kcal / 239 kcal = 677.8

Please read https://sfconservancy.org/copyleft-compliance/principles.htm.... My understanding is that if you do not enforce your copyright (or copyleft in this case) you can lose the copyright.

DNA is made of nucleic acids (deoxyribonucleic acids), not amino acids. DNA is translated to amino acids, except when it encodes long-noncoding RNAs, or miRNAs, or piRNAs, or untranslated regions, or intergenic regions, or transposons, or psuedogenes.

And the protein coding sequences are coded into proteins, except when there are alternative splicing sites, or anti-sense RNAs.

And then the protein sequence folds into their minimum free energy state, except when they are assisted by other proteins, or when they exist as disordered proteins.

And it's not always DNA -> protein. Sometimes proteins make other proteins, for example circular proteins can only be made from other proteins.

Then there are post-translational modifications, which change the RNA sequence between DNA to RNA and RNA to protein. And then there's RNA interference, where miRNAs interfere with RNA to protein translation.

And then there's epigenetics such as DNA methlyation or histone modifications (histones are protein which compress DNA) which change what genes can be expressed when.

Really, for every rule that you've been taught, there is an exception. Biology is so much more complicated than we understand. And understanding how to make a biological computer (different from the biological computing of Adelman, yes the same Adelman as RSA) would involve understanding how all the pieces fit together.