[HN Gopher] YC's founder-matching service helped medical records...
___________________________________________________________________
YC's founder-matching service helped medical records AI startup
Hona
Author : mstats
Score : 32 points
Date : 2024-05-04 18:48 UTC (3 days ago)
(HTM) web link (techcrunch.com)
(TXT) w3m dump (techcrunch.com)
| neilv wrote:
| Why is Birthdate a required field on the User Profile required
| for YC founder-matching?
|
| (Gender is optional.)
|
| Also, once you get past that User Profile, to the founder-
| matching Profile, the very top of the first form says "Your
| profile will get much more attention if you add a picture." Why
| implicitly endorse that, which has undesirable historic baggage
| of prejudice and unfairness?
|
| In hindsight, I realize that the prominence of schools on the
| profiles (and the text field using the example of Stanford)
| didn't jump out at me, maybe because it's common on a US resume
| (unlike gender, birthdate, or photo).
|
| Maybe other people will have additional "what?" reactions to
| other things in the forms.
|
| Edit: On the matching profile form, it won't let you submit the
| form and proceed until you add a photo. ("Please fix the errors
| below before continuing.") I was able to get past that by
| uploading a blank image, but I'm not going to proceed further if
| it seems like current YC thinking isn't a match for me.
| bdw5204 wrote:
| A recent PG essay[0] was revealing about why colleges are so
| prominent on those profiles and more generally about YC's
| thinking:
|
| > At this point I'm going to tell you something you might not
| want to hear. It really matters to do well in your classes,
| even the ones that are just memorization or blathering about
| literature, because you need to do well in your classes to get
| into a good university. And if you want to start a startup you
| should try to get into the best university you can, because
| that's where the best cofounders are. It's also where the best
| employees are.
|
| [0]: https://www.paulgraham.com/google.html
| neilv wrote:
| I have to partly disagree with him there.
|
| My schools included MIT and Brown. And I generally really
| liked the students at both. And I would hire many of them.
|
| But I've also hung out in other circles that attract smart
| people, where I'd selectively hire first, before prioritizing
| random people from the fancy-pants schools.
|
| And I've worked too many places where some of the smartest
| and most accomplished people came from BFE schools.
|
| Also, note that undergrad programs have gotten insane with
| the gaming and the hoop-jumping. Even worse than when it was
| helicopter parents merely sending kids to prep schools, SAT
| prep, and the checkoff extracurriculars (e.g.,
| "volunteering"). And that insanity can select for what you
| _don 't_ want to hire, to do actual work.
|
| If you want "How to Start Google [where everyone focuses on
| the entrance exam, and then on gaming their promotions]" then
| you absolutely should hire Stanford students, who had to
| exercise very similar thinking to get into Stanford at this
| point.
|
| Or, if you want to run a startup investment scam, where you
| need to apply some similar thinking of the _appearance_ of
| accomplishments, and knocking down ducks lined up for you.
|
| But if you want to solve problems, or build something that's
| not just headed for a financial exit, then Stanford-track
| kind of thinking seems probably counterproductive.
|
| Imagine getting a new hire at an early startup, and they
| innately think that their job is to knock off sprint tasks,
| and have other appearances of performance.
|
| And when you try to explain thinking about the organization's
| goals, and that their job is to help the team collectively
| achieve that... they think you're just spewing empty
| platitudes, like they were taught to spew for college
| application personal statements and job interviews.
|
| You would've been so much better off hiring the scrappy
| person from the community college who hasn't had Stanford-
| track BS baked into them since grade school.
| throwaway444441 wrote:
| You sound like the kind of person I'd actually want to
| start a business with. Maybe cofounder matching should be a
| HN feature instead of a tinder clone.
| ra0x3 wrote:
| Wow super insightful comment. Particularly this excerpt:
|
| > If you want "How to Start Google [where everyone focuses
| on the entrance exam, and then on gaming their promotions]"
| then you absolutely should hire Stanford students, who had
| to exercise very similar thinking to get into Stanford at
| this point.
|
| I realize that this applies to Google post-2015, but do you
| think it also applies to early Google? Where you actually
| had to get things done?
| cmrdporcupine wrote:
| First few employees at Google? Stanford yes because
| that's where they got the business/investment
| connections. Clearly, because L&S both came from that
| background as did family, and the whole Susan W's garage
| thing, etc. etc.
|
| After that? Actually building the thing? Not really?
|
| Jeff Dean:
|
| University of Minnesota, B.S. Computer Science and
| Engineering (1990) University of Washington, Ph.D.
| Computer Science (1996)
|
| Tho I guess _Sanjay_ was more "ivy league" (correct me
| where I'm off thing, I'm Canadian and don't understand
| the American obsession with "brand name" on schools):
|
| Cornell University (SB, 1987) MIT (SM, 1990; PhD, 1995)
|
| I guess Urs was Stanford, though.
|
| _But_ I would insist that 1996 is a lot different from
| 2024.
|
| And, honestly, pardon me, but YCombinator isn't funding
| the next Google. They're trying to fund the next Uber or
| whatever. I don't get the impression they have the guts
| to invest in kind of long-play deep tech infrastructure
| like a Google. It all seems like "X but for Y" Web2.0ish
| companies.
| debacle wrote:
| You're comparing a hand picked group from one school to the
| median at another. The most successful student at your
| average community college will likely be more successful
| than the median MIT grad, but that doesn't prove your
| point.
| slashdev wrote:
| Very insightful and accurate take IMHO.
| carabiner wrote:
| Notably, you didn't attend MIT or Brown for undergrad, just
| terminal master's degrees. The selection process for
| undergraduates at those schools is about 100x more rigorous
| than that for MS programs.
| neilv wrote:
| Why notably? I don't think we're communicating well, if
| that's what you latched onto. Are you suggesting I didn't
| also interact with undergrads when I said I liked the
| students there, or is this borderline ad hominem?
|
| Perhaps the meta-relevance is that some people fixate on
| the prestige and exclusivity of these things, and fight
| aggressively to acquire that for themselves, and to
| preserve that status. And some other people help
| perpetuate superiority myths, including in hiring.
| shuangly wrote:
| So I think the statement is more like _statistically_ in
| good universities you 're more likely to meet other great
| engineers and cofounders. It of course doesn't mean all
| students from good universities are better from scrappy
| people community college. When it comes to reality, I'm a
| startup founder and how would I screen people? If someone
| went to Stanford it's a strong good signal. If another guy
| went to community college but has exceptional technical
| skills of course I'd hire him than an average Stanford
| graduates. School or past experience is just one signal but
| it's a strong signal.
| neilv wrote:
| > _When it comes to reality, [...] If someone went to
| Stanford it 's a strong good signal._
|
| I just said why I don't think it's currently a good
| signal, in reality.
|
| Unless the reality is last-decade VC-growth startups,
| where the priority was _appearances_ , from top to
| bottom.
| siva7 wrote:
| So far about 30 startups were founded over the matching service
| out of how many who got into a batch over the years? We're
| talking probably about less than one percent. The service puts
| also a prominent emphasis on formal credentials like which
| university you graduated from. Your idea doesn't matter at all,
| it's just personality and some credentials.
| dtnewman wrote:
| _> Your idea doesn't matter at all, it's just personality and
| some credentials._
|
| One of YC's overriding beliefs is that yes, your idea doesn't
| matter at all.
| ahstilde wrote:
| to be fair, they consider traction over all other credentials
| Zenzero wrote:
| Given my experience with LLMs I am skeptical that enough prompt
| engineering and tuning can adequately summarize patient records.
| I don't expect an LLM to adequately interpret data in a way that
| is useful to me. While much is missed on skimming records, an AI
| summary sounds like a false sense of security.
| mountainriver wrote:
| Yup I actually worked on this problem and we abandoned it
| because the downside of missing important patient information
| was too risky
| mousetree wrote:
| I recently received a 2 page written report of an MRI. I
| couldn't understand a single word of it. While waiting a few
| weeks for my appointment at the next specialist I used ChatGPT
| to explain it and I found it to be very helpful in
| understanding what was going on. I felt it did a better job in
| explaining in lay terms than the specialist doctors did
| (radiologists, orthopedic surgeons, rheumatologists).
| SaberTail wrote:
| How do you know ChatGPT explained it to you correctly? How do
| you know it didn't give you a wrong explanation that sounds
| plausible to a lay person?
| renewiltord wrote:
| For my part I know the same because I did it and then gave
| both the content and the generated report to my father: an
| orthopaedic trauma surgeon / my cousin a radiologist and
| surgeon / my other cousin a cardiovascular surgeon. All for
| different things.
|
| Varying degrees of enthusiasm (my father was the most
| thrilled) but all remarked it was accurate. Of course I did
| these things for a laugh because I can just access the
| appropriate professional on-demand.
| CharlieDigital wrote:
| There are at least two distinct workloads for ChatGPT: 1)
| No context, general knowledge workloads, 2) Contextual
| workloads using either RAG or direct input of context.
|
| _In general_ , GPT will not hallucinate or give bad
| responses on the latter since it's working from a specific
| corpus of information (whether from RAG or from some
| context directly provided by the user). This is, of course,
| not foolproof since some of it is dependent on a good
| prompt and some of it is dependent on the question (GPT is
| well known to be poor at math, for example). But for
| summarization, it is exceedingly good.
|
| The former is where it can tend to hallucinate and generate
| bad responses.
| danenania wrote:
| Something to keep in mind with every AI startup is that they're
| trying to skate to where the puck will be in a year or two. You
| could still be right, but it's worth considering what happens
| to any particular use case if the model gets 2 or 3x smarter.
| hehdhdjehehegwv wrote:
| I've likewise investigated this space, or an adjacent problem.
| TLDR: you can get good results with very high probability.
|
| However, medicine and law are not areas where "probably" even
| at a 95% level is going to cut it. You need 100%.
|
| (Sure a person makes mistakes as well, but we understand that
| and have systems in place to catch and correct. Whereas people
| are just tossing out LLM chatbots that are basically black
| box).
|
| It's very similar to self driving cars, getting to 95% isn't
| too hard now, but that last 5% to not accidentally kill a few
| people here and there is WAY WAY WAY harder.
| CalRobert wrote:
| I find the cofounder matching to be.... fine, but as noted
| elsewhere in these comments don't like the focus on age, photo,
| etc.
| julianeon wrote:
| This actually does shift my view on founder-matching a little, in
| the positive direction.
|
| I wonder if it's best to march into it like someone in a new city
| starting to date. I would guess that it's mostly a numbers game:
| apply to a lot of people you'd be compatible with, wait for their
| responses to winnow it down, then get serious about the handful
| that are left.
| squigglydonut wrote:
| Hey this worked for me this month actually.
|
| I looked through about 800 profiles and did end up finding a co-
| founder after a couple of months.
|
| 800 profiles, 10 matches, 5 zoom calls, 1 accept.
|
| I created a compelling and clear profile and my approach was to
| be as open an honest as possible about what I'm looking for. As
| the technical founder I offered 50/50 split as advised by YC.
|
| I stand by my choice and I know that I trust my co-founder 100%
| and I'm looking forward to the earth shattering product we will
| launch together in August.
___________________________________________________________________
(page generated 2024-05-07 23:01 UTC)