Chemical modifications of DNA are so amazing, and underpin so much DNA related research and engineering. Illumina and Moderna would not exist without DNA mods. It’s very cool that the set of tools is expanding further!
“ Guided by the removable DNA page numbers, Sidewinder achieves an incredibly high fidelity in DNA construction with a measured misconnection rate of just one in one million, a four to five magnitude improvement over all prior techniques whose misconnection rates range from 1-in-10 to 1-in-30.”
I wonder if this is even a problem, since you could amplify the correct sequence with PCR afterward.
I don’t think PCR is necessarily relevant here. I had the impression that this would be lost useful at linking multi-kb fragments together. If we are looking at sizes much above 2kb, PCR is going to struggle to generate full length fragments efficiently.
I didn’t see this technique as having DNA modification per-se, but a novel way to managing the hybridization process. It’s stock (well engineered) oligos, if I read it correctly.
Intuitively I agree some kind of selective amplification should be able to correct for the mistakes. But I think it will be complicated. Because the filtering process needs to be much more complex. It can’t just chemically match to a known subsequence - you won’t know where the mistake might be in a long sequence.
It amplifies sequences that contain the two primer sequences on each end of the target. So if you had synthesized sequence XYZ with some mistakes like YZX, then you could target X and Z and purify.
You're correct that PCR has a limited max length, but it is longer and cheaper than vanilla DNA synthesis.
Are they good at translating scientific jargon specific to a niche within a field? I have no doubt LLMs are excellent at translating well-trodden patterns; I'm a bit suspicious otherwise..
In my experience of using it to translate ML work between English->Spanish|Galician, it seems to literally translate jargon too eagerly, to the point that I have to tell it to maintain specific terms in English to avoid it sounding too weird (for most modern ML jargon there really isn't a Spanish translation).
It seems to me that jargon would tend to be defined in one language and minimally adapted in other languages. So I’d not sure that would be much of a concern.
I would look at non-English research papers along with the English ones in my field and the more jargon and just plain numbers and equations there were, the more I could get out of it without much further translation.
> Should I be able to sue McDonalds if I let my kid eat 100 of them in one sitting?
There are other options for addressing social problems besides lawsuits. Other rich places in this world are not nearly as fat as us. I suspect environments also matter for social media addiction. We should investigate why!
It's actually because they have more lawsuits and more severe lawsuits, leading companies to be afraid of breaking the law so they don't, and then lawsuits decrease.
Lawsuits are the one official mechanism for righting wrongs. They're the only mechanism that the perpetrator of a wrong can't just choose to ignore.
The downside of opsec is that it breeds paranoia and fear about legal, civic participation. In a way, bullshit investigations like this are an intimidation tactic. What are they going to find - a bunch of Minnesotans that were mad about state-backed killings?
One metric you could use is how often publications are mentioned by patents, and how often those patents lead to economic value. By this metric, it is valuable.
The number of PhDs we have is currently too many given the amount of money we have for project grants. But there is no evidence that the money we allocate to research is too large. If anything, you could argue the opposite.
I would be delighted if the private market funded basic research - the seed ideas that lead to patents.
People have tabulated the value of the academic pipeline, from grant to paper to patent to stock valuation. It is overall very valuable, even if you grant the very real issues with our hyper-competitive grant system.
This intellectual capital is valuable, despite whatever the latest populist memes about professors claim.
I am also skeptical (despite having 0 faith in the new owners). However, I am a bit confused: why would new ownership alone cause technical issues? It seems like they set new requirements that required new software. Even if the reqs are content-agnostic, I am curious what they are and how they differ from the previous tiktok.
Data migrations, new staff permissions and policies, merging AWS or other cloud accounts and their complex IAM policies, enrolling devices into new corporate networks, Okta setup, corporate firewalls. There are hundreds of reasons that moving to a new corporate ownership can cause technical problems.
All of these will be happening as well as hundreds of other moving pieces (domain name transfers, DNS changes, adjusting compute resources to maximise existing contracts with providers, migrating metrics systems impacting alerts and infra scaling etc etc etc). It's impossible to say from the outside which of these are causing issues.
> I thought they were already using usa-based cloud infrastructure
Unless they were already using the same provider on the same account with the same IAM policies as their new owner (they were not, obviously) this is irrelevant.
I think the fairest test is: what is the best and fastest way to reduce medical uncertainty? For rare ailments with a single cause and exclusive symptoms, that can be accurately described with simple language (no medical jargon), its possible that an LLM is better than a doctor.
For more ambiguous situations where you need actual tests, I am skeptical of using LLMs.
“ Guided by the removable DNA page numbers, Sidewinder achieves an incredibly high fidelity in DNA construction with a measured misconnection rate of just one in one million, a four to five magnitude improvement over all prior techniques whose misconnection rates range from 1-in-10 to 1-in-30.”
I wonder if this is even a problem, since you could amplify the correct sequence with PCR afterward.
reply