5 ms·

I might have looked positively at this if the bureaucrats would have actual skin in the game and would be judged by the outcome in 5-10 years. But in the likely

I might have looked positively at this if the bureaucrats would have actual skin in the game and would be judged by the outcome in 5-10 years. But in the likely case that this will fail and most of the money will be drained through lobbyism etc., nobody will talk about it and they will come up with the next thing they can throw money at to come across as being super innovative. Disclaimer: I used to work at a company that was particularly good in draining these kind of funds without delivering useful outputs.

singhrac 6y ago
Maybe you could use your experience to propose a system (whether funding vehicle restrictions or proper incentives) for that not to happen in this case? I think we’d all be excited by that. How can we avoid spending money on lobbyists? Maybe the structure should be more VC-like?
- rusk 6y ago
  No point arguing with ideology.
  - CryptoPunk 6y ago
    It is arguably quite ideological to insist that a large monopolistic bureaucracy, or a conglomeration of them - e.g. the EU - will be efficient in allocating vast amounts of resources when basic game and microeconomic theory suggests otherwise. Still, an argument for state subsidies, despite some inevitable degree of inefficiency, can be made.
    - rusk 6y ago
      Point was that despite repeated successes with this style of investment the libhurdurians still insist it’s a bad idea, totally blind to the fact that lassaiz affaire is just as prone to degeneration. It’s ideology as much as any other approach and when you find people insisting on it without any concrete examples it’s just plain ideology you’re dealing with and there’s just no point wasting your time.
      - CryptoPunk 6y ago
        I'm a libhurdurian and I don't think it's necessarily a bad thing. A public goods argument can be made for strategic investments. I also don't think it's been conclusively shown that there have been "repeated successes with this style of investment". Demonstrating that would require knowing what the opportunity cost of the spending for these projects was, which is impossible. Claiming it's an open-and-shut case that these kinds of investments will yield positive results is just as ideological as you accuse dissenters of being.
        rusk 6y ago
        I’m not claiming it’s open and shut. Every time something like this comes around you have all these guys giving out. I don’t say free market doesn’t work in all cases but just that it doesn’t work in all cases. Libertarianism is a dead end. Also BSD Unix.
        CryptoPunk 6y ago
        Your earlier statement, that "lassaiz affaire is just as prone to degeneration" is just as ideological as anything you responded to. Neither of you have proven your case. It's entirely possible that you are right, and there's a factual basis for your assertion. It's also entirely possible that the party you are in opposition to is right, and there's a factual basis for theirs. Neither point has been supported in this discussion with anything beyond rhetoric.
- lostmsu 6y ago
  An option to try is making a mixed merit+vote system instead of just voting.
- hndudette2 6y ago
  Putting structure aside for now (although I agree we want an answer to that), I don't see why we even want governments allocating capital to projects like this in the first place when the space is already very competitive with loads of private money. It seems like a classic case of big government doing big government things to justify their own existence.
- jabl 6y ago
  "More VC like" has been proposed by Mariana Mazzucato. If the state invests in a startup/spinoff, the state should get equity in return rather than just some lofty words about future jobs or such. Not for the purpose of creating long time state owned companies, but for cashing out when/if the investment is successful, just like VC's. This would both provide the state with money, removing the need to raise that same money via taxes, and hopefully also better align the incentives of the funding organisation by providing measurable goals.
yostrovs 6y ago
If it does succeed, it will be a perpetual, repeatedly brought up, example of how the EU can get things done. The other projects will remain forever forgotten.
zackmorris 6y ago
Government spending isn't the problem, it's who gets the money. In the US I see it all going to established players. Which is fine for stuff like finding a COVID-19 vaccine. But is a total disaster for promoting innovation. I have a computer engineering degree and have watched nothing of interest happen in semiconductor manufacturing for over 20 years. Yes we have blue LEDs now. Yes we have fast video cards. But if you look at single-threaded performance per clock cycle, it's been stuck within the same order of magnitude since the Pentium got 20 stage instruction pipelines and huge caches (which in hindsight were big blows for SIMD and MIMD). It would take a paltry amount of money, in the low millions of dollars, to design real (general purpose) multicore processors that deliver orders of magnitude better performance than what we have today. But unfortunately lawmakers are for the most part technologically illiterate, and technologists are unduly skeptical of any type of programming outside the mainstream. So my dream of a sub-$1000, 1000+ core CPU with a modest amount of RAM per core (between 1 MB and 1 GB) that can be programmed with existing tools like Erlang/Go/MATLAB/Julia and even Docker is just never gonna happen. And without that, there is no viable road to really experiment with stuff like AI, physics simulations etc without renting time in the cloud. We have the impression that progress is being made on these endeavors today, but things look a little different to me, watching them play out at a glacial pace, at mind boggling expense, over 3-4 decades. I mourn what might have been.
- alokrai 6y ago
  If the money required "to design real (general purpose) multicore processors that deliver orders of magnitude better performance than what we have today" is in low millions of dollars and technical challenge isn't too daunting, why haven't VCs invested in such ventures?
- baryphonic 6y ago
  > It would take a paltry amount of money, in the low millions of dollars, to design real (general purpose) multicore processors that deliver orders of magnitude better performance than what we have today. But unfortunately lawmakers are for the most part technologically illiterate, and technologists are unduly skeptical of any type of programming outside the mainstream. > > So my dream of a sub-$1000, 1000+ core CPU with a modest amount of RAM per core (between 1 MB and 1 GB) that can be programmed with existing tools like Erlang/Go/MATLAB/Julia and even Docker is just never gonna happen. And without that, there is no viable road to really experiment with stuff like AI, physics simulations etc without renting time in the cloud. We have the impression that progress is being made on these endeavors today, but things look a little different to me, watching them play out at a glacial pace, at mind boggling expense, over 3-4 decades. I mourn what might have been. I'm really curious what specifically you mean by this. I see similar issues on the software side (my degree is in CS). Software is incredibly bloated and horrible at interoperability. Unix had pipes back in the early 70s, and somehow with GUIs and then mobile "apps," we've regressed. Identity-based security has failed time and again. And only rarely has software design progressed meaningfully beyond structured programming from the late 70s (not to mention the languages). Moore's law has given software developers around over 100M x improvement over the past 40 years, yet ordinary people would scarcely notice. TL;DR: I see a lot of the same flaws in software, dominated by fads and popularity only of what is mainstream. I'd really like to hear a dissident point of view from the hardware side.

singhrac 6y ago

Maybe you could use your experience to propose a system (whether funding vehicle restrictions or proper incentives) for that not to happen in this case? I think we’d all be excited by that. How can we avoid spending money on lobbyists? Maybe the structure should be more VC-like?

rusk 6y ago
No point arguing with ideology.
- CryptoPunk 6y ago
  It is arguably quite ideological to insist that a large monopolistic bureaucracy, or a conglomeration of them - e.g. the EU - will be efficient in allocating vast amounts of resources when basic game and microeconomic theory suggests otherwise. Still, an argument for state subsidies, despite some inevitable degree of inefficiency, can be made.
  - rusk 6y ago
    Point was that despite repeated successes with this style of investment the libhurdurians still insist it’s a bad idea, totally blind to the fact that lassaiz affaire is just as prone to degeneration. It’s ideology as much as any other approach and when you find people insisting on it without any concrete examples it’s just plain ideology you’re dealing with and there’s just no point wasting your time.
    - CryptoPunk 6y ago
      I'm a libhurdurian and I don't think it's necessarily a bad thing. A public goods argument can be made for strategic investments. I also don't think it's been conclusively shown that there have been "repeated successes with this style of investment". Demonstrating that would require knowing what the opportunity cost of the spending for these projects was, which is impossible. Claiming it's an open-and-shut case that these kinds of investments will yield positive results is just as ideological as you accuse dissenters of being.
      - rusk 6y ago
        I’m not claiming it’s open and shut. Every time something like this comes around you have all these guys giving out. I don’t say free market doesn’t work in all cases but just that it doesn’t work in all cases. Libertarianism is a dead end. Also BSD Unix.
        CryptoPunk 6y ago
        Your earlier statement, that "lassaiz affaire is just as prone to degeneration" is just as ideological as anything you responded to. Neither of you have proven your case. It's entirely possible that you are right, and there's a factual basis for your assertion. It's also entirely possible that the party you are in opposition to is right, and there's a factual basis for theirs. Neither point has been supported in this discussion with anything beyond rhetoric.
lostmsu 6y ago
An option to try is making a mixed merit+vote system instead of just voting.
hndudette2 6y ago
Putting structure aside for now (although I agree we want an answer to that), I don't see why we even want governments allocating capital to projects like this in the first place when the space is already very competitive with loads of private money. It seems like a classic case of big government doing big government things to justify their own existence.
jabl 6y ago
"More VC like" has been proposed by Mariana Mazzucato. If the state invests in a startup/spinoff, the state should get equity in return rather than just some lofty words about future jobs or such. Not for the purpose of creating long time state owned companies, but for cashing out when/if the investment is successful, just like VC's. This would both provide the state with money, removing the need to raise that same money via taxes, and hopefully also better align the incentives of the funding organisation by providing measurable goals.

yostrovs 6y ago

If it does succeed, it will be a perpetual, repeatedly brought up, example of how the EU can get things done. The other projects will remain forever forgotten.

zackmorris 6y ago

Government spending isn't the problem, it's who gets the money. In the US I see it all going to established players. Which is fine for stuff like finding a COVID-19 vaccine. But is a total disaster for promoting innovation. I have a computer engineering degree and have watched nothing of interest happen in semiconductor manufacturing for over 20 years. Yes we have blue LEDs now. Yes we have fast video cards. But if you look at single-threaded performance per clock cycle, it's been stuck within the same order of magnitude since the Pentium got 20 stage instruction pipelines and huge caches (which in hindsight were big blows for SIMD and MIMD). It would take a paltry amount of money, in the low millions of dollars, to design real (general purpose) multicore processors that deliver orders of magnitude better performance than what we have today. But unfortunately lawmakers are for the most part technologically illiterate, and technologists are unduly skeptical of any type of programming outside the mainstream. So my dream of a sub-$1000, 1000+ core CPU with a modest amount of RAM per core (between 1 MB and 1 GB) that can be programmed with existing tools like Erlang/Go/MATLAB/Julia and even Docker is just never gonna happen. And without that, there is no viable road to really experiment with stuff like AI, physics simulations etc without renting time in the cloud. We have the impression that progress is being made on these endeavors today, but things look a little different to me, watching them play out at a glacial pace, at mind boggling expense, over 3-4 decades. I mourn what might have been.

alokrai 6y ago
If the money required "to design real (general purpose) multicore processors that deliver orders of magnitude better performance than what we have today" is in low millions of dollars and technical challenge isn't too daunting, why haven't VCs invested in such ventures?
baryphonic 6y ago
> It would take a paltry amount of money, in the low millions of dollars, to design real (general purpose) multicore processors that deliver orders of magnitude better performance than what we have today. But unfortunately lawmakers are for the most part technologically illiterate, and technologists are unduly skeptical of any type of programming outside the mainstream. > > So my dream of a sub-$1000, 1000+ core CPU with a modest amount of RAM per core (between 1 MB and 1 GB) that can be programmed with existing tools like Erlang/Go/MATLAB/Julia and even Docker is just never gonna happen. And without that, there is no viable road to really experiment with stuff like AI, physics simulations etc without renting time in the cloud. We have the impression that progress is being made on these endeavors today, but things look a little different to me, watching them play out at a glacial pace, at mind boggling expense, over 3-4 decades. I mourn what might have been. I'm really curious what specifically you mean by this. I see similar issues on the software side (my degree is in CS). Software is incredibly bloated and horrible at interoperability. Unix had pipes back in the early 70s, and somehow with GUIs and then mobile "apps," we've regressed. Identity-based security has failed time and again. And only rarely has software design progressed meaningfully beyond structured programming from the late 70s (not to mention the languages). Moore's law has given software developers around over 100M x improvement over the past 40 years, yet ordinary people would scarcely notice. TL;DR: I see a lot of the same flaws in software, dominated by fads and popularity only of what is mainstream. I'd really like to hear a dissident point of view from the hardware side.