cottonseed

You cannot bruteforce this. Exhibiting a unknotting of K with n moves only gives you an upper bound u(K) <= n. Proving u(K) = n is an entirely different matter.

Those are the same. To see that, just flip over L before performing the connect sum.

Proteins are linear molecules consisting of sequences of (mostly) 20 amino acids. You can see the list of amino acids here: https://en.wikipedia.org/wiki/Amino_acid#Table_of_standard_a.... There is a standard encoding of amino acids using single letters, A for alanine, etc. Earlier versions of ESM (I haven't read the ESM3 paper yet) uses one token per amino acid, plus a few control tokens (beginning of sequence, end of sequence, class token, mask, etc.) Earlier versions of ESM were BERT-style models focused on understanding, not GPT-style generative models.

I actually prototyped a system like this, mostly as an exercise to learn about crypto. You can't feasibly host or verify proofs on-chain, so you need external trusted verifiers (e.g. oracles). Making sure the oracles can't front-run proof submission is a challenge. Standard formal proof system (like Lean) are sufficiently expressive, although they weren't built for this and need to be modified to make sure a proof hasn't introduced any additional axioms, as you note. The proof system also becomes a point of attack, so you'd probably want multiple, independent verifiers (which themselves have been formally proved correct). I believe these exist for some proof systems, although I'm not sure about Lean's kernel.

Ultimately, I don't think this is really practical, and investing in AI proof agents is the way to go.

2001: A Space Odyssey (1968)

Yes. EleutherAI is doing it, probably one of many:

https://www.eleuther.ai/projects/gpt-neox/ https://github.com/EleutherAI/gpt-neox https://arxiv.org/abs/2204.06745

They have a 20B parameter model. I think the primary dataset for these open models is The Pile: https://arxiv.org/abs/2101.00027 (web scrape, pubmed, arxiv, github, wikipedia, etc. There is a nice diagram on page 2 that summarizes the contents.)

AllSpice might be what you're looking for?

AllSpice: A git platform for hardware engineers

https://www.allspice.io/

Where are you losing people in your hiring process? Are you getting no initial applications? Do you give the coding exercise but they never do it? Do you give an offer but they turn it down? There is a lot of speculation in this thread but you should have the data.

Mathematician here. What do you want this for? Even if you had them, you probably wouldn't understand the definitions anyway.

As others say, there is no standard, and conventions vary by subfield, publication, author and over time. This is esp. true at the research level, where the mathematical content is still being worked out. Subfields have certain conventions, and well-written books and papers will normally introduction notation or include an index of notation, esp. if the notation is novel or they different from the usual conventions. You could start compiling something like this by going through the standard undergrad and grad textbooks for each subject.

One of the best technologists I ever worked with denied his interest in technology until he was around your age. He was a professor at a top school in CS and started some innovative and impactful companies. I quit my job at 34 to study math and got my PhD at 40. After that, I left math to work in biology and I run a data science/engineering group at a premier biology research institute. I will probably change things up again before I'm done. I am not unique, there are many examples of this:

https://mathoverflow.net/questions/7120/too-old-for-advanced... https://math.stackexchange.com/questions/237002/too-old-to-s...

You are young and life is long. Go do what you love.

edit: My email is in my profile. Reach out if you want to chat.

I thought this recent Hackaday article did a good job putting the current Chernobyl situation in perspective: https://hackaday.com/2021/05/14/increased-neutron-levels-at-...

edit: submitted: https://news.ycombinator.com/item?id=27179475

It is pretty clear Jim Keller did something pretty remarkable at Apple and then AMD (I know less about his work at Tesla). I tried to dig into the stuff he's said and written to understand what he did and how he did it. Say what you want about Fridman's interview style, that interview was probably the most insightful thing I found.

You guys are getting upvotes?

Hail at the Broad Institute of MIT and Harvard | Software Engineer | Boston, MA | ONSITE, https://hail.is, https://broadinstitute.org

The Broad Institute of MIT and Harvard was launched in 2004 to improve human health by using genomics to advance our understanding of the biology and treatment of human disease, and to help lay the groundwork for a new generation of therapies.

The Hail team's mission is to build tools to enable rapid analysis and exploration of biological datasets (100s of TB and tripling yearly). We are committed to open science and everything we do is open source. We currently develop in Python, Scala/Java, and C/C++ and use Spark, Kubernetes, Google Cloud Platform (GCP) and AWS, but will use any tools we need to get the job done. Come help us build the future of big scientific data analysis.

We have two positions:

Update: The Site Reliability Engineer position has been filled.

We also have a front-end/designer position that will be posted shortly. Email below, get in touch if you're interested.

You don't need experience in biology or our particular technologies. We work in a highly multi-disciplinary environment (with software engineers, biologists, bioinformaticians, doctors, operations, statisticians, etc.) Self-improvement is a fundamental part of our culture. You must be excited to be challenged and learn new things.

I'm the hiring manager. Get in touch with me directly if you have any questions: cseed@broadinstitute.org.

You can learn more about the project here: https://hail.is, https://github.com/hail-is/hail

We are one of several software engineer groups at the Broad that are hiring. You can find more positions here: https://broadinstitute.wd1.myworkdayjobs.com/broad_institute

I think it's going to largely depend on build times and how much infrastructure you need to spin up during during tests.

Some discussion for the same question in a recent thread: https://news.ycombinator.com/item?id=21679714

We're quite a bit smaller but have similar numbers: 15-20m right now. We're dominated by build time (build caching might help) and schlepping docker images.

You might like 5 Elements of Effective Thinking, too.

Check out What the Best College Teachers Do and What The Best College Students Do by Ken Bain.

The Broad Institute:

https://www.broadinstitute.org/about-us

https://broadinstitute.wd1.myworkdayjobs.com/broad_institute

I work there. My group builds scalable tools for genomic data analysis:

https://hail.is

We're about to post two job reqs, for an SRE and front-end/design position. Email in my profile. Get in touch if you're interested.

gnomAD is the largest public dataset of human genetic variation:

https://gnomad.broadinstitute.org/

They recently a 7 paper collection in Nature: https://www.nature.com/collections/afbgiddede. They're also hiring an SRE:

https://broadinstitute.wd1.myworkdayjobs.com/en-US/broad_ins...

Lots of other jobs at various levels throughout the institute. Biology knowledge generally note required (I had none), although it helps (but be prepared to learn).

For context, a skim of random internet sources claim that TSMC produced 1.1M 7nm wafers last year.

I don't think so, but you need to manage/sync the infrastructure state yourself.

college

The author specifically mentions college friends. Sounds like they were only friends with CS majors even then.

It might be useful if you say what you're trying to become an expert in. Ask a vague question, get a vague answer.

always come out frustrated

You can't stop there.

Yes, a lot of expert knowledge is locked up in the heads of experts. It is very hard (if not impossible) to write down all the implicit and explicit knowledge that experts have, so it doesn't always happen. It's very hard to become an expert alone. I think this also says something about the nature of expertise: it is something that is constructed by experts themselves in their minds. There was a story that a famous mathematician would tell is grad students, holding up an important book, "You should know everything in this book ... but don't read it!"

Function composition is always associative.

You're still confused. There is no function composition here.

op in the example above is just some other function, like +. The associativity of + and function composition are true for totally unrelated reasons. Associativity of plus is an inductive argument that follows from the Peano axioms.

Function composition says:

(f o g) o h = f o (g o h)

as functions. It is true because unary function application "serializes" function applications. Formally, I mean:

    ((f o g) o h)(x)
      = f(g(h(x)))
      = (f o (g o h))(x)

The Domain and Codomain for both functions are exactly the same.

This is not true.

def List[A].foldLeft[B](z: B)(op: (B, A) => B): B

def List[A].foldRight[B](z: B)(op: (A, B) => B): B

Notice the signature of the fold op: the arguments types are swapped. This is because fold left and right on a list [a, b], say, is the difference between:

(z op a) op b

and

a op (b op z)

(If this isn't compelling enough, consider [a, b, c].) Not all functions are associative. For example, consider a cryptographic hash function.

Maybe watch the talk first before commenting? All these questions are answered in the talk.

Google wants to create an open, innovative ecosystem for silicon so it will be easier for them to build accelerators for their workloads to meet the growing demand for compute. TPU is only one example of the kind of accelerators they want to build. Tim directly addresses this in the talk: https://youtu.be/EczW2IWdnOM?t=407.

I don't think so. Tim explains in the talk, designs must be submitted via a public Github repository. I think the whole point is to create an open ecosystem.

Yes, there are lots of open-source RISC-V cores. Tim Edwards of efabless has another talk about creating a RISC-V based ASIC SOC: https://www.youtube.com/watch?v=EsEcLZc0RO8 based on PicoRV: https://github.com/cliffordwolf/picorv32. PicoRV is part of the efabless IP offerings. The chips will have a PicoRV harness on them.

efabless already runs the shuttle service that Google and efabless are going to fund here. From the efabless home page:

$70K, 20 WEEKS, 100 SAMPLES

Note quite $50K, but close.

Not sure why you're being downvoted. I haven't been following this and I appreciated the summary.