4 ms·

This argument assumes that "moving fast enough to make mistakes" is actually getting us closer to a usable technology. There's not much reason to think this is

This argument assumes that "moving fast enough to make mistakes" is actually getting us closer to a usable technology. There's not much reason to think this is the case based on the opinions of other experts in the field. Theranos also "moved fast" - straight into a ditch.

api 4y ago
Theranos had other problems, like a fundamentally untenable design that many biologists criticized as being physically impossible. Not hard, but physically impossible. Nevertheless it is true that moving fast increases risk. You either fail spectacularly or succeed spectacularly.
- tovej 4y ago
  or barely succeed and have to make up the debt to technical design and process quality later, taking longer in the end than if you had built a solid foundation for your endeavor
- AlexandrB 4y ago
  Some of the things Musk has claimed are probably physically impossible using the device Neuralink has constructed. For example, there is indeed a treatment for depression that involves stimulating parts of the brain using electrodes. But the parts you need to reach are way in there[1]. Neuralink's device is not capable of reaching these areas in its current form. However I'm way out of my depth on this. The best I can do is listen to skepticism and try to evaluate it on its merits vs. the things that Musk says and his track record of over-promising. [1] https://www.shockmd.com/2008/01/20/6-different-locations-for-deep-brain-stimulation-in-depression/ https://www.shockmd.com/2008/01/20/6-different-locations-for...
  - nvrspyx 4y ago
    I won't speak at all on Neuralink, but you don't necessarily need to stimulate that deep into the brain directly. You can piggyback off existing functional connectivity between that deep area and an area closer to the cortical surface. Functional connectivity is when one area sends activation or inhibitory signals to another area when stimulated. This is the mechanism behind using transcranial magnetic stimulation (TMS) to treat depression ([0] as just one example) despite TMS's shallow stimulation depth. Using the functional connectivity also provides an avenue of sending inhibitory signals, not just activation signals. With that said, it requires functional imaging (fMRI) to determine the functional connections. Functional connectivity also isn't indefinitely static and it does change over time. . 0: https://ajp.psychiatryonline.org/doi/full/10.1176/appi.ajp.2019.19070720 https://ajp.psychiatryonline.org/doi/full/10.1176/appi.ajp.2...
    - nvrspyx 4y ago
      I should add that it's much more nuanced and complicated than how I've explained. Along with determining if a functional connection is excitatory or inhibitory, you need to ensure its direction as well because functional connections aren't necessarily bidirectional. You also want to be confident in how stimulating the target endpoint will affect the functional connection over time. It may potentially strengthen it (think "training") or weaken it (think "desensitizing"), thus potentially affecting the efficacy of the stimulation over time. As with most scientific fields, knowledge is built up one step at a time and you need to stand upon the shoulders of prior work before producing an application with confidence.
- pclmulqdq 4y ago
  Once Theranos's patents lost their teeth, devices similar to what was in those patents appeared on the market within a few years. All from big pharma companies. Today, a lot of blood testing machines can run many tests on a few hundred microliters of blood. When you go get blood work done, they collect a vial, but often use only a few drops at a time.
  - pvaldes 4y ago
    After being concentrated?
    - pclmulqdq 4y ago
      Usually not, actually, although some more complicated tests do involve sample prep. For most tests, it's a small microfluidic cartridge and a tiny amount of blood.
      - pvaldes 4y ago
        I would remain skeptical about this claim. There are a lot of exploratory tests that can be done with a drop of blood and have a solid history behind, for more complicated tests we would fall again in a "Theranos magic" case. I hope that the investors finally learn their lesson and accepted that if you want to produce good results you need to starting with good data. In science there is not sense in finding the minimum algorithm possible that vomits the desired results. Alive beings are complicated So either the claim that somebody is building the nbext theranos machine or recycling theranos machines for doing what theranos couldn't... either is false, or is just old technology wrapped in glitter, or would send serious doubts in the confidence in the results. It depends on the details.
SketchySeaBeast 4y ago
> The mistakes leading to unnecessary animal deaths included one instance in 2021 when 25 out of 60 pigs in a study had devices that were the wrong size implanted in their heads Especially when it's sloppy mistakes that don't provide scientific value.
mindslight 4y ago
> Theranos also "moved fast" - straight into a ditch There's no need to make a figurative analogy to Theranos when Musk runs another company with the same development philosophy, with a product that will literally drive straight into a ditch. Move fast and break things was edgy precisely because it was going against longstanding of common sense. Shaking things up a bit worked for what was fundamentally a low stakes software project (a CRUD microblogging app), where the users weren't even stakeholders. Back in physical world of hardware engineering, it's a terribly sloppy philosophy.
- concordDance 4y ago
  It worked for SpaceX.
  - mindslight 4y ago
    I haven't followed SpaceX enough to know if they operate(d) that way, or if it worked reasonably. If it did, I'd guess it was due to increased guardrails. At first glance, you'd think the design of rockets would be more critical than cars. But the issue in Tesla's case isn't the trickiness of the technical problem, but rather the ethics of testing on consumers in populated areas. Taking it as a given that a rocket can't be launched over say a major city (one "guardrail"), it would seem to matter a lot less if a bunch of unmanned rockets blew up on the pad or over a test range.
microtherion 4y ago
In addition to the technological question, there also seems to be a personality risk here. The kind of people who thrive in an environment where speed and marketable results are prized above anything else, and standards of care are neglected are unlikely to deliver top notch results.
concordDance 4y ago
> There's not much reason to think this is the case based on the opinions of other experts in the field. There is however good reason to think this is the case by looking at the history of many other developments, such as planes (so many crashes and deaths in the 1900s) and rockets (many SpaceX booms).
woeirua 4y ago
Yep, you can't rush good science. Neuralink will just be another company on the trash heap of failed ideas at the rate they're going right now.

api 4y ago

Theranos had other problems, like a fundamentally untenable design that many biologists criticized as being physically impossible. Not hard, but physically impossible. Nevertheless it is true that moving fast increases risk. You either fail spectacularly or succeed spectacularly.

tovej 4y ago
or barely succeed and have to make up the debt to technical design and process quality later, taking longer in the end than if you had built a solid foundation for your endeavor
AlexandrB 4y ago
Some of the things Musk has claimed are probably physically impossible using the device Neuralink has constructed. For example, there is indeed a treatment for depression that involves stimulating parts of the brain using electrodes. But the parts you need to reach are way in there[1]. Neuralink's device is not capable of reaching these areas in its current form. However I'm way out of my depth on this. The best I can do is listen to skepticism and try to evaluate it on its merits vs. the things that Musk says and his track record of over-promising. [1] https://www.shockmd.com/2008/01/20/6-different-locations-for-deep-brain-stimulation-in-depression/ https://www.shockmd.com/2008/01/20/6-different-locations-for...
- nvrspyx 4y ago
  I won't speak at all on Neuralink, but you don't necessarily need to stimulate that deep into the brain directly. You can piggyback off existing functional connectivity between that deep area and an area closer to the cortical surface. Functional connectivity is when one area sends activation or inhibitory signals to another area when stimulated. This is the mechanism behind using transcranial magnetic stimulation (TMS) to treat depression ([0] as just one example) despite TMS's shallow stimulation depth. Using the functional connectivity also provides an avenue of sending inhibitory signals, not just activation signals. With that said, it requires functional imaging (fMRI) to determine the functional connections. Functional connectivity also isn't indefinitely static and it does change over time. . 0: https://ajp.psychiatryonline.org/doi/full/10.1176/appi.ajp.2019.19070720 https://ajp.psychiatryonline.org/doi/full/10.1176/appi.ajp.2...
  - nvrspyx 4y ago
    I should add that it's much more nuanced and complicated than how I've explained. Along with determining if a functional connection is excitatory or inhibitory, you need to ensure its direction as well because functional connections aren't necessarily bidirectional. You also want to be confident in how stimulating the target endpoint will affect the functional connection over time. It may potentially strengthen it (think "training") or weaken it (think "desensitizing"), thus potentially affecting the efficacy of the stimulation over time. As with most scientific fields, knowledge is built up one step at a time and you need to stand upon the shoulders of prior work before producing an application with confidence.
pclmulqdq 4y ago
Once Theranos's patents lost their teeth, devices similar to what was in those patents appeared on the market within a few years. All from big pharma companies. Today, a lot of blood testing machines can run many tests on a few hundred microliters of blood. When you go get blood work done, they collect a vial, but often use only a few drops at a time.
- pvaldes 4y ago
  After being concentrated?
  - pclmulqdq 4y ago
    Usually not, actually, although some more complicated tests do involve sample prep. For most tests, it's a small microfluidic cartridge and a tiny amount of blood.
    - pvaldes 4y ago
      I would remain skeptical about this claim. There are a lot of exploratory tests that can be done with a drop of blood and have a solid history behind, for more complicated tests we would fall again in a "Theranos magic" case. I hope that the investors finally learn their lesson and accepted that if you want to produce good results you need to starting with good data. In science there is not sense in finding the minimum algorithm possible that vomits the desired results. Alive beings are complicated So either the claim that somebody is building the nbext theranos machine or recycling theranos machines for doing what theranos couldn't... either is false, or is just old technology wrapped in glitter, or would send serious doubts in the confidence in the results. It depends on the details.

SketchySeaBeast 4y ago

> The mistakes leading to unnecessary animal deaths included one instance in 2021 when 25 out of 60 pigs in a study had devices that were the wrong size implanted in their heads Especially when it's sloppy mistakes that don't provide scientific value.

mindslight 4y ago

> Theranos also "moved fast" - straight into a ditch There's no need to make a figurative analogy to Theranos when Musk runs another company with the same development philosophy, with a product that will literally drive straight into a ditch. Move fast and break things was edgy precisely because it was going against longstanding of common sense. Shaking things up a bit worked for what was fundamentally a low stakes software project (a CRUD microblogging app), where the users weren't even stakeholders. Back in physical world of hardware engineering, it's a terribly sloppy philosophy.

concordDance 4y ago
It worked for SpaceX.
- mindslight 4y ago
  I haven't followed SpaceX enough to know if they operate(d) that way, or if it worked reasonably. If it did, I'd guess it was due to increased guardrails. At first glance, you'd think the design of rockets would be more critical than cars. But the issue in Tesla's case isn't the trickiness of the technical problem, but rather the ethics of testing on consumers in populated areas. Taking it as a given that a rocket can't be launched over say a major city (one "guardrail"), it would seem to matter a lot less if a bunch of unmanned rockets blew up on the pad or over a test range.

microtherion 4y ago

In addition to the technological question, there also seems to be a personality risk here. The kind of people who thrive in an environment where speed and marketable results are prized above anything else, and standards of care are neglected are unlikely to deliver top notch results.

concordDance 4y ago

> There's not much reason to think this is the case based on the opinions of other experts in the field. There is however good reason to think this is the case by looking at the history of many other developments, such as planes (so many crashes and deaths in the 1900s) and rockets (many SpaceX booms).

woeirua 4y ago

Yep, you can't rush good science. Neuralink will just be another company on the trash heap of failed ideas at the rate they're going right now.