I have a question for anyone in an R&D/Bell Labs-esque place - are there any good recommendations for similar places to work, particularly outside the US? "Old" Google apparently was, but going by what ex-googlers have been saying it hasn't been the case for a long while now.
If you are in the UK Deepmind seemed to be that way to me when I applied. A lot of research groups doing hard research in a wide array of fields, cutting edge AI stuff, a core team of SEs developing in house programs for researchers, and research engineers for productionization. And no, I didn't get the job lol.
I've heard mixed things about it as a company but GResearch (also UK) seemed like an interesting R&D software / math mix in the vein of investment banking. I applied there over ten years ago so YMMV at this point, who knows.
I work at a very successful HFT. It’s a special place, but we’re not advancing the state of the art in fundamental science like the Bell Labs/Microsoft Researches of the world.
It is pretty close to zero. Has anyone thought about how much a trade of a broadly traded security will cost them in recent times, or ever thought it would not happen near instantly?
Thanks for your suggestions! I'm currently in The Netherlands but I'll keep these in mind (though they sound more "math-ey" than engineering-ey but that's fine).
>though they sound more "math-ey" than engineering-ey
Isn't math necessary for most real-world engineering, and even in CS research? And what were you expecting? You asked for a cutting edge Bell-Labs type of R&D place, and that's what research is all about, even in CS.
You'll work with a lot of new yet-unproven theoretical concepts for which you need a lot of math to prove they have a high chance of working in practice and being better than existing solutions, before someone approves budget for the costly development and implementation of an actual product.
It really depends on what you're looking at. Disclaimer, I'm a mechanical engineer and not a CS guy. A lot of the stuff if you're doing say quantum computing is understandably math heavy. But if you're say designing a flying kite-like generator, it's more "engineering math", if you know what I mean? (Which is what I'm comfortable with)
Any serious place will only be looking for folks that are bonafide math geniuses in addition to their actual specialty.
In the case of mechanical engineers, maybe 1 in 50 bonafide geniuses in mechanical engineering are also simultaneously math geniuses. Just my personal hunch.
There's really not that many positions, so it's unrealistic to expect recruiting standards to be much lower.
If I remember correctly, MSR worked on a practical implementation of Code-Contracts for C# which incorporated the (all-important) compile-time verification of method preconditions, postconditions, and class invariants (without the need for hand-written refinement-types, which is how we do things today): as I understand it, the
compile-time part of system could support any assertion represented as a pure-function - think of it as C#'s take on Ada's assertions, improved tenfold, and it even shipped for a now-unsupported older version of C# and .NET: https://learn.microsoft.com/en-us/dotnet/framework/debug-tra...
Had Microsoft put more backing behind it, then C# could present itself as a language to supplant Ada in safety-critical applications, and replace C/C++ in other applications.
I have hope the feature will come back one-day - there are whole slews of bugs that can be eliminated (such as when passing EF entity types around with unintentionally null member-properties).
Disney Imagineering? I have no direct experience but I'm continually impressed by the technical innovations they make public and the effects they manage to achieve in their products.
Worked for them for about 10 years. Fun place, but I'm not sure it'll ever be like the old days where research was done for research's sake. So many things are now off-the-shelf, and not much is invented in house.
between westchester and nyc one has a lot of varying environments to live in if working at IBM Research. I used to commute up to it from Manhattan without any problem (nice reverse commute in general). Hawthorne (now closed) was a bit easier than Yorktown, but both were fairly easy (if memory serves me correct, yorktown was only an additional 15 minutes or so).
Thanks, product design is something I've also considered. I'd guess Apple/MS style companies are probably still the best bet, even though you're more restricted than what it probably was at Bell.
I'm not sure how many, if any, are doing blue sky research (vs product-directed "research") any more the way that Bell Labs, IBM Watson and Xerox Parc used to do.
Look at what's going on with ML/AI - DeepMind now merged with Google Brain seemingly with a product focus, FAIR now moved into a product group alongside Meta's GenAI group, Microsoft essentially outsourcing AI to OpenAI, OpenAI may as well call itself GPTCorp - a single-product commercial enterprise.
I guess it's not surprising given how short term the thinking is of today's publicly traded companies.
The reason Bell Telephone built that organization is that they had a government-sanctioned monopoly on local phone service, and to justify the continuation of this, they wanted to find ways to show they were contributing to society.
Could someone like Google or Microsoft build a Bell Labs today? Yes, almost certainly, but there's no financial incentive to do so. And the shareholders would not be pleased if you told them you were going to spend their money on something with no connection to the business.
A bigger question for the present is: why are the universities failing so badly? Their incentives have not changed, but we don't get the kind of innovative research we got out of Bell Labs. I don't know what the answer to that is.
Agreed. The role of monopoly profits in funding pure research almost can’t be overstated. Similarly, the geographic monopolies of newspapers (before the Internet) funded quality journalism.
"A bigger question for the present is: why are the universities failing so badly? Their incentives have not changed, but we don't get the kind of innovative research we got out of Bell Labs."
In my opinion, professors at research universities have to contend with the pressures of raising grant money in a competitive environment combined with "publish or perish" pressures. Even post-tenure there are ways universities could punish "non-productive" faculty members at institutions where professors are regularly expected to publish at top journals/conferences and raise grant money. It's not that much different from the pressure their corporate research counterparts face, where they have to regularly justify their employment by producing a regular flow of research results that have business impact. This pressure to produce results on a regular schedule, whether in industry or in academia, is something that I strongly believe stifles science and forces scientists to make evolutionary "sure bets" instead of working on riskier, more revolutionary projects like the ones that Bell Labs and Xerox PARC researchers got to work on during those labs' heydays.
Alan Kay, who worked at Xerox PARC in the 1970s, has a lot to say about supporting long-term, revolutionary research here (http://worrydream.com/2017-12-30-alan/).
In my opinion, the simple answer is that we need institutions that provide researchers the freedom and space to work on riskier, longer-term projects, and we also need funding to support such research.
For my personal career, I'm reminded of this quote from physicist J. J. Thomson made over a century ago:
"Granting the importance of this pioneering research, how can it best be promoted? The method of direct endowment will not work, for if you pay a man a salary for doing research, he and you will want to have something to point to at the end of the year to show that the money has not been wasted. In promising work of the highest class, however, results do not come in this regular fashion, in fact years may pass without any tangible results being obtained, and the position of the paid worker would be very embarrassing and he would naturally take to work on a lower, or at any rate a different plane where he could be sure of getting year by year tangible results which would justify his salary. The position is this: You want this kind of research, but, if you pay a man to do it, it will drive him to research of a different kind. The only thing to do is to pay him for doing something else and give him enough leisure to do research for the love of it." (from https://archive.org/details/b29932208/page/198/mode/2up).
My career goal is to find a way to make a living outside of research that provides me enough time to do the research I want to do. This way I'm free from the pressures of either "publish or perish" from academia or "profit or perish" from industry.
Eh, I don't think today's scene is altogether that different from back then. In every case you have a research organization tied to an immensely profitable main enterprise. The vast majority of the work force works on the "product" side and only a small number of researchers are doing blue sky stuff.
This describes Bell Labs and Xerox Parc, as well as modern counterparts like MSR and DeepMind. As always, only a very small portion of the work force gets to do blue sky stuff - the rest have to do the "mundane" bits of making money.
Let's not be fooled by rose-tinted glasses here - even in its heyday Bell Labs was small fraction of the overall Bell operations, and likewise Xerox Parc an extremely prestigious but yet small slice of the overall enterprise.
> DeepMind now merged with Google Brain seemingly with a product focus, FAI
I don’t really agree that training massive causal LMs is a “product focus”.
I agree that there is an increasing product focus in orgs like OAI, but a lot of that is coming from new growth rather than trading off with base research.
> I don’t really agree that training massive causal LMs is a “product focus”.
I'd hope that Google DeepMind's mission statement is a bit broader than that, but certainly Gemini seems to be the focus. It seems to me there's a world of difference between DeepMind's original goal of developing AGI (via whatever means - as a research objective), and now being told they have to build LLMs.
If we compare Google to Meta, it seems it used to be that DeepMind was equivalent to FAIR as a pure research organization, and Brain equivalent to Meta's product focused ML group(s), but now the joint Google DeepMind is more akin to Meta's GenAI group, and there is no unfettered research group at Google left free to pursue AGI in any way other than hoping it can be developed out of LLMs. However, FAIR also seems constrained now that they have been moved into a product-focused part of the organization (under CPO Chris Cox).
Did you work in this field? I just don’t find that a fair characterization. They are all still doing blue-sky research.
There is a ton of further research to be done in deep models. A lot of it will incorporate LLMs because that is currently the most powerful primitive you have.
Research is a lot more closed now but I would not take that to mean research is no longer happening.
No - I don't work in the ML field, or for any of these companies (but I have built a Torch-like NN framework in C++ from scratch, and followed the early transformer development closely, so do understand the tech).
I'm just writing about the changes to these organizations, and their corporate governance, as reported in the press. I don't think you need to be an insider to appreciate the difference between, say, DeepMind as an independent entity pursuing AGI anyway they saw fit (RL), and now as part of Google DeepMind apparently tasked with developing SOTA LLMs. No doubt this is still a research vs pure engineering endeavor, but hard to call it blue sky when the research direction and goal is so proscribed. I personally don't believe that LLMs (or RL for that matter) are the path to AGI, but at least DeepMind used to have the flexibility to pivot and pursue whatever lines of research they felt were most promising. Do they still have that flexibility today?
> I have built a Torch-like NN framework in C++ from scratch
Mm, care to share? I am skeptical when people make claims like this, even though it is very achievable. Many people are simply "playing house" when it comes to ML tech.
I would not believe everything you read even in the tech aligned press, it is very often false. Google Deepmind is not exclusively researching LLMs.
I never released my framework, and don't intend to (abandoned this a good few years ago), but it was more than "playing house" ... It was complete enough to build/train a convnet that worked with CIFAR-10, supporting both GPU via cuDNN and CPU via my own Tensor class with MKL/IPP BLAS/etc acceleration. The API was Torch-like where you build the graph (create nodes, then connect them), then run it. I was in process of writing a version 2 with support for RNNs and auto migration of tensors from CPU to/from GPU, but gave up since PyTorch had since appeared (obviously a better approach) and it became increasingly obvious how ridiculous it was for a one-man project to attempt to catch up to SOTA!
Well, if you're an insider at one of these companies, then please enlighten us!
What has changed from DeepMind as independent entity to Google DeepMind (following the hivemind LLM "path" to AGI ;) ) ?
What does it mean for FAIR to now be part of a product group ?
I'm sure the CEOs of these companies hope these changes are significant and will better align these research organizations with their corporate goals and marketplace opportunities ...
Is OpenAI pursuing AGI-directed research outside of LLMs, or is it now all hands on deck towards next GPT 5.0 product release ?
Google Deepmind is still publishing tons of non-LLM research. The difference is that these directions are being incorporated as Gcloud products faster.
I am not sure about FAIR, but they have been chatbot/LLM/ASR focusd from before this change.
OpenAI has definitely pivoted towards product. !
Most of that pivot is through new growth while still maintaining existing research products, but I wont deny there is a new level of product focus.
As a freelance product designer I'm always learning and always excited about work. There are so many interesting companies working on so many interesting problems.
It depends on the needs of the client, and Ive done a little bit of everything from designing moulds and sourcing factories for physical products, to writing frontend code to help ship.
