After reading everything there, I am left with more question than when I opened the page. Is the page honest?
Who is giving the money?
Why are they sponsoring people?
Only a handful of questions to preapply? Why? There used to be pages like this only to get people's emails in a target audience for spam purposes, is this is?
I can apply, I have an amazing idea to pitch, but it looks more shady than legitimate, sadly.
I hate to say it...but I agree. That really pains me to say because I wish there were opportunities like this beyond a typical PhD program. The concept is great. But this doesn't seem like the best implementation of that concept.
It does feel like there are more questions than answers here. What are the differences between full "members" and "affiliates"? Why aren't there more recognizable names behind this? I don't see a celebrity research behind this, so to speak, and that is precisely the kind of person I think you should get to start this kind of organization.
I also don't understand what "decentralized Bell Labs" is supposed to mean. The monetary funding is clearly centralized under one organization, so do they just mean the researchers won't literally be physically colocated? Is that meaningful enough a differentiater to highlight it? I can't think of a single industrial lab that has researchers who work in isolation with mentorship here and there. Didn't we try this with the Thiel Fellowship already?
However I don't think this is shady, and I think that's an uncharitable way to interpret it. It looks like the members' credentials are legitimate. The problem I have is that there doesn't seem to be a top researcher with tour de force accomplishments backing it. That probably sounds prejudiced, but I consider it a negative signal that there are no such researchers who were even willing to affiliate their names, let alone their guidance to the project.
Founder here. We have secured partial funding but not all that we intend to raise. This is just a pre-application - an improved website and application will be released later. If you submit your application now we'll follow up with you when we release our full application or convert it automatically.
We're focused on creating a world-class mentorship community before increasing our Fellow outreach efforts.
it was posted earlier, but never made it on the top 10 pages for reasons we couldn't figure out despite getting plenty of upvotes on /newest. To check if it was actually alive we posted a couple dummy comments after a while, and then abandoned the post.
(disclosure, I'm a mentor and the post was sent to a few mentors, which likely tripped some bot checker)
That's why I called it naughty and not blatant manipulation. Now if this were an ICO or a company looking for an easy flip in the overhyped AI market, it would be different.
It looks like the founder was just trying to kick off discussion without being explicit about her involvement. The submitter (Aaron Lin) is also part of the organization.
In the scheme of things, I give them a pass. Raising attention for an initiative like this can be challenging. It's not actually dishonest.
The mentorship pool is mostly focused on AI and related fields, and it's high leverage for relatively few resources invested (check out https://aigrant.org/)
Some perspective: the existing, Federal competition in this space is the NSF GRFP, which is about $165k over five years, instead of $100k for one year. Obviously if you win the Shannon Fellowship for two years, you are coming out ahead of the GRFP. The GRFP is also extremely competitive. It does open some academic doors to you though, since you are paying for your own PhD basically, advisors don't have to worry about how to pay you, and your ideas had merit to the NSF so there are two positive signals for you already.
Like some other comment said, science is all about collaboration. Not being subject to "publish or perish" is nice but a lot of social capital has accumulated around the black hole of anguish that is the research infrastructure, it's hard to work with people that aren't on the paper treadmill because even if you get paid whether or not you publish, your collaborators don't.
From experience, in research, one year is not a long time. It will go by much faster than you think. If you have an idea and you need a short runway to just get started, that sounds about right. Doing anything with human systems seems perilous though, you can spend months just figuring out how to get lab space and how to recruit subjects and setting up equipment (oh yeah, I guess this $100k is also your equipment and space budget). Also from experience, in research, $100k will go faster than you think.
It would be awesome to win this and whoever is doing this deserves major credit for stepping up and doing this, but don't get too starry-eyed when you're thinking about what you could do. If you've got a project already waiting, "facing downhill" and it needs a push, this sounds like a good opportunity. If you want to chase a "deep idea" and only have a hazy idea of where to go, this is probably not what you want.
I think it’s an open question how the internet will support research collaboration. There’s clear success stories such as Wikipedia when it comes to mass scale contribution, but fewer specific research innovations outside of open source projects. My best guess is that the types of collaborations will simply be different (The Bazaar and The Cathedral documents this well).
The internet is arguably better at flourishing novel ideas than Bell Labs was. For instance, machine learning Twitter is an active and relatively small community. Excellent ideas from obscure researchers often rise to prominence within hours when retweeted by a more followed researcher. This wouldn’t happen as often in a building, and the type of idea discovery that Twitter allows is just the tip of the iceberg. I think it’s worth realizing the significant downsides to in person collaboration (nothing is saved, it's entirely flow vs stock, you need enough status to be hired in the first place).
I was in the first class of the Thiel Fellowship, a decentralized program that tended to centralize itself anyway in pockets (groups of us rented big houses and lived together). The combination of in-person and distributed was a great midway, and I predict the Shannon Fellowship will be similar.
I'm excited to see what the Bell Labs of the Internet looks like, and I think it's too early to say decentralization will simply not work.
Wikipedia and the like are fine when each contribution is usually quite independent.
AI pursuit often involves a singular potential breakthrough idea that's larger than one person can work on in isolation, for any reasonable time frame. Certainly a group of people with low living expenses might be able to run a team of 3-4 people with this sort of budget, but it sounds like it's intended for individual researchers.
It also worked because researchers had long-term employment with ample research funding and long periods of time to think outside of the rapid-pace of today's culture in academic and industry labs. Projects such as UNIX, C, C++, and S were built by a small set of bright people who had a significant portion of their time to devote to a single project. They did user-testing, reimplementing, and even had the luxury of starting-over from scratch several times to get the first stable prototype ready.
Something like the proposed model doesn't work if they are only giving someone "1 year"; they'll spend a good chunk of that time looking for the next job or source of funding.
The fact that Joscha Bach is involved gives this more credibility than anything else.
I'm tempted to apply but I'm guessing the work I want to do is too far afield of traditional ML work, maybe someone can tell me if that's wrong.
The work I want to do is not in iterations on algos, like I've been working on with CV problems, but better defining intelligence and building testing mechanisms for it along the lines of the work of the Anytime Intelligence Test [1].
Too little is being done on baselining real world intelligence and so what we get are closed world, fully observable tests, which are impressive, but aren't really functionally useful when we're trying to build or test AI's generalization capabilities.
You do need a Google account to get to the pre-application page, just FYI.
(So I'm out. Also, reading about "unconventional" researchers being sought out is kind of disappointing compared to this likely unintentional requirement.)
Might be worth changing the title to note the subject limitations. "Shannon Labs – $100,000 Fellowship to support independent intelligence researchers" works.
Parnian (in comments) is a friend of mine and the submitted piece requires some disucssion. I'm just happy it raised these questions. OFC these questions are good to answer, and if this succeeds it'll be cool. I'm not sure what the program will emerge as - still in flux.
Most of what other comments bring up is [IIRC from talking about this]
The website is intended as a fixed figment for the current goals of shannon labs
seems like a nice thing to have to show investors
1. funding
- 1 funded, website is tool
- plan for 10 to alpha
1a. Only 100k?
- funding spreads this way
- about enough for a year in pure software
- could be more but start and grow
2. motivation
- More research is good!
- Unsure about equity free :(
- In a position to secure funding for a cool project
3. relationship to academia
- I'm sure many of these will materialize with academic connections
3a. publishing req?
- I'm not sure most people will write a paper and submit to a journal. I've seen a lot of beautiful JS that has been put together as a research artifact. I like this?
If the motivations behind this are as altruistic as they're made out to be, props to them. This is EXACTLY what applied healthcare AI researchers/developers need. Granted, $100K for just one year isn't enough runway for an end-to-end solution, but I suspect you could get enough of a ball rolling to pick up sponsors or investors.
Most of the work that needs to be done in this space is NOT sexy. Building cool AI models is by far the easiest part of the pipeline, even considering the fact that data acquisition can be difficult. No one wants to get into the weeds of regulations, hospital protocols, EHR integrations, or HCI issues pertaining to medical providers. You don't get publications from sifting through this mess. So most researchers don't touch it. Additionally, the slow development cycles in healthcare tech are unattractive to many industry developers. There's a graveyard filled with failed startups and company initiatives that expected to jump into healthcare with their awesome tech only to get crushed by at least one of the multiple walls that shield the industry from innovation.
I personally find myself caught between a rock and a hard place when it comes to healthcare + AI innovation. If I am being completely honest, I care more about building a useful product than publishing research papers, and I care more about improving patient care than making a bajillion dollars. This puts me in a weird spot in terms of choosing between an academic vs industry route. My current strategy is to find a way to take the healthcare innovation I want to do and repackage it as PhD research.
I'd tots apply for this, assuming there's no massive catch. :)
Hi there, I work at OpenAI. This isn't related to OpenAI in any official sense, though there are some people involved who either have had links to us in the past, or currently have links to us.
For those who aren't familiar, Scott Gray (one of the mentors) is a complete genius. Two years after he left Nervana his neural network kernels (which AFAIK haven't been updated since) are still the fastests available: https://github.com/soumith/convnet-benchmarks
I'm sure the people behind this have good intentions and the initiative could work out, but it's also a little bit bizarre that they have only secured "partial" funding yet are offering to throw 100k at underfunded ideas.
- it'll be easier to persuade investors to put in money when there are good applications (then "unconventional background" applications will not have big impact here, people will probably still look at the conventional criteria)
- with some similar program exist already (aigrants), 100k is needed to compete with other program, to attract applicants, to grab your attention in news title...
I'd assume that two basic things required for such an endeavour are A. to actually have the money you want to give out and B. to have the trust of the community.
Since they admit that they don't have the money, and neither their website nor the less than honest comment by the "founder" (of what? for now, a website) inspire any trust, it seems like this whole thing can safely be ignored.
The best approach, as often in these cases, is a bona fide private conversation. Just a very short pitch of your idea and a timetable for the 12 months would suffice, imho, and make clear immediately where you do stand and them.
disclaimer: I know Parni personally, and she did not ask me to chime in here.
Parni's intentions are honest, even if her program is a little bit ill defined at this stage.
I think she's started this fellowship because she'd observed several of her friends struggling to find an avenue to pursue their research direction. Realistically, it's financially difficult for many to pursue a ML PhD when they can be earning fat stacks in industry. I think that programs like this help balance out the incentives, for engineers who need normal income. Programs like OpenAI and Google Brain aren't the last word here, the more the merrier!
Personally, I'd prefer to see a program which sets up some sort of equity in an investment fund for ML PhD students. Academia is still probably the best way to push the frontiers of public knowledge, but its almost insane to forgo salary & equity comp for 5 years in a high growth field. Making the opportunity cost of a PhD lower seems like a good outcome.
RE: the question of if there is funding or not, my best guess is that if there is a good applicant pool it would be highly unlikely the program go unfunded.
Despite being in academia, it's an interesting idea, though I wonder about the time period (what do you do after that year?) and honestly...I've spent enough time on the internet that "Independent Researcher" makes me twitch involuntarily.
Don’t think they have figured out what to do after the first year or whether this turns out as a failed investment for investors ... like lost new gigs turn out to be.
These are not really good fields to target unless you’re planning to fund tons of psychedelics/chemical research (which is probably illegal) under the guise/goal of intelligence augmentation. There is tons of research in “AI” etc. Most of the ideas are bad, it’s not like there are ideas that are getting overlooked.
Who is giving the money?
Why are they sponsoring people?
Only a handful of questions to preapply? Why? There used to be pages like this only to get people's emails in a target audience for spam purposes, is this is?
I can apply, I have an amazing idea to pitch, but it looks more shady than legitimate, sadly.