biotechbio

I think an important comparison is the LLM with the supervised training data in context. I wonder how the LLM would perform if you provided all, or a representative subset of, the data used to fine tune the CNN in the prompt and then asked it to classify a new case.

This strategy is how LLM-like models have achieved leading performance on tabular prediction tasks (TabPFN).

Hello fellow palo alto person

Is this headline misleading? From reading the article, Berkeley is pausing authorizations for Course Credit CPT, which enables international students on visas to work in exchange for credits to complete their degree programs. It doesn’t outright ban work experience if it is an “integral part of an established curriculum”.

The headline reads like Berkeley is suspending authorization of international student workers in general, but this actually seems like a relatively narrow restriction?

Not true, there is a growing class of "operator" VCs that both fund and help run early stage companies.

The obvious other factors in this reframing of equity compensation effectively render it a useless take.

The disparity in expected value calculation is due to assuming you can predict the future after joining a company, which is obviously not true. If you can always tell a company is going to succeed after 1 year of working there, you should leave, as you are clearly destined to be the most successful venture capitalist in history

Not to be overly negative, I think this perspective is somewhat misleading as it lets one rationalize valuing startup equity as anything more than a gamble. Which it is, and always has been. Sometimes gambles work out though.

This post is a little long winded. At times I agreed, and others I felt the author was doing a little too much hand-wringing on what-is-mine vs. what-is-the-AI's.

The opportunity is and always has been the possibility of accelerating work. Honestly, if something works (I mean genuinely, actually works) I don't care at all about what craftsmanship or insight went into its creation.

We value these things because they have become correlated with quality. We now have the opportunity to decouple these things; maybe something that took no effort will be just as good as a painstaking human labor.

The risk is if this doesn't come true. If we let our skills degrade and get ahead of our skis, embracing "slop" that superficially appears to "work", we will eventually pay the price. Financially and culturally, it seems like we are already all-in on the bet that it will work.

I hope it does, I just want to solve the problems I am working on.

When you use a microscope to magnify something, the objective (magnifying lens) is literally taking the Fourier transform of the image. The optical system recovers up to a limiting frequency, determining the spatial resolution of the image.

I disagree, in my opinion passing the bar exam is necessary but not nearly sufficient for competently practicing law.

The bar is an imperfect filter. One could study for the exam and pass and still be hugely deficient in ability as an attorney.

I would argue there's no exam that could replace the evaluative and experiential component of 3 years in law school, and accreditation helps enforce at least some standard of quality in the profession. More incompetent lawyers -> more wasteful behavior -> a more bloated and slower legal system -> worse outcomes for everyone.

I think reducing barriers to completing the legal education (part-time programs, lower cost, etc) are better avenues for increasing access.

Some thoughts on this as someone working on circulating-tumor DNA for the last decade or so:

- Sure, cancer can develop years before diagnosis. Pre-cancerous clones harboring somatic mutations can exist for decades before transformation into malignant disease.

- The eternal challenge in ctDNA is achieving a "useful" sensitivity and specificity. For example, imagine you take some of your blood, extract the DNA floating in the plasma, hybrid-capture enrich for DNA in cancer driver genes, sequence super deep, call variants, do some filtering to remove noise and whatnot, and then you find some low allelic fraction mutations in TP53. What can you do about this? I don't know. Many of us have background somatic mutations speckled throughout our body as we age. Over age ~50, most of us are liable to have some kind of pre-cancerous clones in the esophagus, prostate, or blood (due to CHIP). Many of the popular MCED tests (e.g. Grail's Galleri) use signals other than mutations (e.g. methylation status) to improve this sensitivity / specificity profile, but I'm not convinced its actually good enough to be useful at the population level.

- The cost-effectiveness of most follow on screening is not viable for the given sensitivity-specificity profile of MCED assays (Grail would disagree). To achieve this, we would need things like downstream screening to be drastically cheaper, or possibly a tiered non-invasive screening strategy with increasing specificity to be viable (e.g. Harbinger Health).

This is a pretty cool study with some interesting findings! Cancer immunotherapy has a long history but has become very prominent in recent years. (Fun fact: the senior author on this paper, Ed Engleman, co-founded of one of the first cancer cell therapy companies, Dendreon, in the early 90s). However, the success of immunotherapies has been limited by the immune-exclusionary nature of the tumor microenvironment (TME). Why some tumors are immune-hot and others are immune-cold is still a very open research question.

In this study, the authors demonstrate pretty convincingly that erythropoietin (EPO, a hormone that stimulates red blood cell production in the bone marrow) reduces the recruitment of tumor-cell-killing T cells to the TME. It does this by acting on tumor macrophages, another type of immune cell, and changes the state of these cells to facilitate accumulation of immunosuppressive cells.

They work out the mechanism largely through mouse models and associative analysis in human tissue samples, but I thought it was interesting that this finding aligns with the clinical observation that cancer patients who receive recombinant EPO for treatment of anemia frequently experience tumor progression.

After reading this, I am going back to check out EPO expression in old datasets that I worked with haha.

Strongly disagree. My son is going to be born into a county with an active measles outbreak caused entirely by misinformation and stupidity. We can’t vaccinate until 6 months. Absolutely preventable. Zero arguments against vaccination for measles. Public health is everyone’s problem.