Hacker News Reader

tsbischofon Ask HN: Employers, why do you want us back in the office?

Part of the in-person experience is the opportunity for people to correct mistakes in communication. Many people are not effective writers, and so pushing for async or writing-heavy communication is difficult at best. Having one-on-ones by voice or video often helps cut down misunderstandings, and in-person is a way to help reduce the barrier to entry for such interactions.

tsbischofon Ask HN: Employers, why do you want us back in the office?

I want people to be around each other enough that they are comfortable interacting. Some of my most painful experiences in management come from having to mediate a low-level dispute between people who work together but never interact on anything other than work. They are quite often on the same page but have different ways of expressing themselves that can cause friction. This often happens as part of having a highly interdisciplinary and international team. It is incredible how much a few shared lunches or coffee breaks can relax people and help them cohere.

I see in-person work as one of the only opportunities for scheduled casual interaction. Mandatory fun time is not a real solution, but the fact that exists often speaks to the perception that casual interaction is an essential part of building a team.

tsbischofon Ask HN: Has anyone worked at the US National Labs before?

I worked at LBNL, at the Molecular Foundry. The day to day was a mix of typical nanoscience work (chemical synthesis, electron microscopy, etc) and work in support of the user facility. In my case that involved consulting on projects involving our users (design of high-throughput screens, teaching spectroscopy, etc), setting up and maintaining instrumentation, and developing workflows for our chemical synthesis robots.

I liked the work and really enjoyed getting to be a consultant on many projects. Turnover is massive among the researchers because there are few permanent positions, and most groups are heavy on postdocs since graduate students tended to be primarily on campus (UC Berkeley).

If pay is a concern, look closely for the open databases of salaries. At LBNL there is the "book of tears" at the library under the cafeteria, listing every employee and their salary. The exact amount you get varies wildly with the department: prior to unionization in 2016, the range was from 20k to 125k annual salary for postdocs. I hear they raised the floor to NIH levels at least, but I assume they did not make NERSC take a paycut.

tsbischofon My PhD qualifying exam was a nightmare but I'm not letting it define me

Hard disagree on dropping quals. Quals were one of the only times where we were forced challenged to reexamine our ideas from a reasonably outside perspective, both to see that we actually knew the material and that we could work outside our little niche. This also happens in the group, at conferences, etc but quals are one of the only institutional opportunities to enforce outside perspective.

Part of the point of the PhD is learning how to approach problems that have no established solution, and someone who lacks breadth or depth will likely find it challenging to be a professional researcher.

tsbischofon My PhD qualifying exam was a nightmare but I'm not letting it define me

Exactly, the PhD is a kind of participation trophy at some point. Quals are the last simple way to decide that someone will not get a PhD.

tsbischofon Ask HN: What is your opinion of “unlimited” PTO?

I recently had to dig into this. Federalism is wild: https://www.paycor.com/resource-center/articles/pto-payout-l...

tsbischofon Ask HN: Where to get help with a hardware idea?

I am glad to chat, my contact is my username at Gmail. I have broad experience in chemistry and materials science, and can at least help you get going in the right direction.

tsbischofon Classifying Python virtual environment workflows

Not all Python users run containers, especially in the sciences. When you are teaching basics of data manipulation to people that have barely touched any programming language before, getting people to grok even virtual environments is not trivial.

tsbischofon Ask HN: Is there academic research on software fragility?

Part of the issue is that there is "no cost" to changing something in the software stack, while there are very tangible costs and barrier to entry to modifying a physical structure. This tends towards a more conservative culture, where controls are placed on designs and changes. This culture also exists in safety-critical software, like that used in medical devices, aviation, and industrial automation.

Otherwise, robustness is relatively expensive because it requires the organization to value the long-term quality and function of the system, at the expense of short-term velocity and malleability. If you are competing with others who can hack together an MVP with 90% functionality overnight, then waiting for the engineered product may be problematic.

tsbischofon Why are clinical trials so expensive? Tales from the beast’s belly

Any particular issues? I am most opposed to the rigid 30-day cycles and the need to spend one just to get on the agenda. But in terms of the questions asked and the due diligence required, I have not found our IRBs to be onerous

tsbischofon Why are clinical trials so expensive? Tales from the beast’s belly

For such a large effect it is quite possible to implement an adaptive trial design with unbalanced arms and interim analysis for efficacy. But you have to ask for this, the FDA will not necessarily suggest it directly.

tsbischofon Ask HN: Firing an employee under a month before vest?

It matters how it is perceived. If everyone involved sees a company firing a bad employee, then good. But if I were in that situation, I would see a company that decided to be greedy and did not care about the sacrifices you made for them.

I treat equity as zero cash value, but that makes it worse because the company is now clawing back something worthless!

That said, I find the whole concept of vesting cliffs to be a bit of nonsense. Vesting I can understand, but why set these cliffs when they are not explicitly tied to regularly-scheduled reviews, performance, or some other checkpoint?

tsbischofon NYU Chemistry Professor Fired After Students Said His Class Was Too Hard

Maybe? Weeder courses in CS based on SICP have similar dropout rates, and the push towards making the curriculum easier causes endless hand-wringing. We could argue that organic chemistry is less central to medicine than recursion and functional programming are to software development, but there is a similar "click" that is a reasonable signal of whether you are likely to parse more complicated concepts.

tsbischofon NYU Chemistry Professor Fired After Students Said His Class Was Too Hard

Part of it is to ensure that your medical students are not terrified by "chemicals". Plus, doing organic chemistry instills into you the idea that reactions may not be pure, and that enantiomers are both real and potentially problematic (see thalidomide). It is at minimum a way to get you to be able to ask critical questions when needed, even if there is already an infrastructure to answer those questions for you.

It is sort of like learning C but programming solely in Python: you can get your job done, but when one of your abstractions breaks it is worthwhile to be able to at least identify the root cause and talk to the right people.

tsbischofon Retraction of Roomtemperature superconductivity in a carbonaceous sulfur hydride

Duplication of work prior to publication is standard in many experimental fields, like synthetic chemistry. In some cases the cost of an experiment is high enough to be a problem, but conductivity measurements are not really that exotic.

tsbischofon Nightdrive

A 256-byte music demo: https://linusakesson.net/scene/a-mind-is-born/

tsbischofon Ask HN: Do you regret being a generalist?

No, from the perspective that it has allowed me to take on interesting projects that were otherwise impossible. In my case, by linking tools from different fields (spectroscopy, computer vision, chemistry, medicine, human systems hacking, ...) and getting something viable together. Once we demonstrate viability and need, we then build a team of specialists to flesh out the details.

Yes, in that that you have to sell yourself carefully. "I solve any problem" is not generally reassuring, because it is always an approximation, and because a specialist might see the pitfalls of a particular approach earlier. Given time you will become a specialist in whatever tool, but that needs to be built into projects or your professional development budget.

tsbischofon Nano-nonsense: 25 years of charlatanry (2010)

Somewhat. At this point nano, 2D materials, topological materials, and other buzzwords of the past are either mainstream or dying. AI/ML for automated experiment design in high-throughput screening is the new hot stuff in experimental science.

tsbischofon Breaking the silence around academic bullying

https://en.m.wikipedia.org/wiki/Suicide_of_Jason_Altom

One of three in a short period

tsbischofon Ask HN: How do you critically evaluate scientific papers?

You can traverse citations to that paper, looking for anything which actually talks about those results.

There are some tools which do this. scite.ai has proven okay for quickly finding some of these citations, sometimes. Nothing beats manually digging in, though.

tsbischofon Ask HN: How do you critically evaluate scientific papers?

Depends on the field a bit, but my general process involves using a few questions to signal one way or the other.

1. What are the actual claims made by the authors? That is, what do they claim to have found in the study? If you cannot find these, there is a good chance the paper is not particularly useful.

2. For each claim, what were the experiments that led to that decision? Do those logically make sense?

3. Are the data available, either raw or processed? Try reproducing one of their more critical figures. Pay close attention for any jumps in series (e.g. time suddenly goes backwards -> did they splice multiple runs?), dropped data points, or fitting ranges. If the data are not available, consider asking for them. If the code exists, read through it. Does the code do anything weird which was not mentioned in the paper?

4. How do they validate their methods? Do they perform the appropriate control experiments, randomization, blinding, etc? If the methods are so common that validation is understood to be performed (e.g. blanking a UV-Vis spectrum), look at their data to find artifacts that would arise due to improper validation (e.g. a UV-Vis spectrum that is negative).

5. Do they have a clear separation of train and test / exploration and validation phases of their project? If there is no clear attempt to validate the hypothesis for new samples, there is a good chance the idea does not transfer.

tsbischofon What a $500k grant proposal looks like

US National Labs (e.g. LBNL) can easily top 200%, based on what my colleagues told me at the time. I never applied for funding myself while there, so I cannot directly confirm the actual figures.

tsbischofon Some argue that synthetic data can make AI systems better

It's also about the issue that people will bias against generating the "difficult" data, even if only subconsciously. Real-world validation is essential to reveal whether you have actually worked through the full phase space.

tsbischofon We sound like idiots when we talk about technical debt

It really should be on some form of balance sheet, in the form of "costs to implement new feature" (debt gets repaid in part by the next customer) or "costs to maintain" (ongoing interest payments that come out of overhead). The big issue with throwing around terms relating to complexity without assigning projected value or probability distribution for that value is that it makes any discussion asymmetric. As someone who knows something about the processes involved but does not have deep knowledge of the code base, it is impossible for me to accurately predict the tradeoffs in discussions. So when we do have to build project plans and discuss tradeoffs, I have to either rely on the people doing the work to build one-off projections or otherwise make things up the best I can based on general patterns.

tsbischofon Ask HN: How to raise funds for rare disease research?

Out of curiosity, what business model were they trying? A few years back we were looking into virtual trial site companies (e.g. Science 37), and the pitch seemed to make sense at the time (focus on studies which require minimal site visits, use telemedicine for routine visits, coordinate with local labs and pharmacies as needed). That structure might be a good fit for rare patient cohorts.

tsbischofon Royal Society cautions against censorship of scientific misinformation online

There is a lot less fakery in science than poorly-designed studies, misleading endpoints, underdocumented or incorrectly documented methods, and cargo-culting. The success of the process comes from having a good filtration process to sift through this body of work, and the idea that there will always be some people in the system doing actually good work.

That said, I have also witnessed plenty of low-level fraud: changing of dates to match documentation, discarding "outlier" samples without justification or even documentation, etc. Definitely enough to totally invalidate a result in some cases.

tsbischofon Arxiv.org reaches a milestone and a reckoning

In principle, submission is viable at many venues. PNAS has a few mechanisms for having established authors endorse your submission, for example. The main challenge is going to be the appearance of credibility to get past the starting point, since many non-affiliated submissions tend to be unpublishable without serious rework, at least in lab-based disciplines.

tsbischofon Arxiv.org reaches a milestone and a reckoning

Part of the challenge is the reputation issue. Academics want to publish where there is prestige, which means there has to be a perception of the venue as good. This typically either comes from reasonable gatekeeping and high-quality peers (e.g. arXiv) or pushing metrics (e.g. Nature Publishing Group journals). Generating an attractive journal is non-trivial and few people with the appropriate knowledge and expertise take the plunge would do so without some guarantee of profit.

tsbischofon A curated list of warez and piracy links

Fix the service problem of actually being able to find films.

Back in the day, Netflix's DVD library was close enough to complete that there was little need to look elsewhere, and their streaming selection was not terrible when it launched. Now that the whole system is fractured, the simple act of finding out who streams your movie is a chore, and in many cases getting it in the original language with proper subtitles is non-trivial.

Create a single point of entry to a high-quality library with minimal friction, and you eliminate the desire to pirate in many people.

tsbischofon Ask HN: Is it even worth reading news outside of HN?

What you miss out on is anything to do with your geographic community. Try looking for sources of local journalism like alt weeklies and zines based in your area. These tend to be published less often and curate for long-running issues rather than attention-grabbing headlines