4 ms·

Regarding computation, I respectfully disagree. I believe the shift from be from software back to hardware again. Failures in Computation: 1) general artificia

Regarding computation, I respectfully disagree. I believe the shift from be from software back to hardware again. Failures in Computation: 1) general artificial intelligence, this has been touted that it's just around the corner since 1970's; so far most practical AI has been just variants of expert systems. 2) Bioinformatics, this was touted in the early 2000's and by the Kurzeweil-crowd as the singularity moment that will bring advances to biotech; unfortunately, all of the mountains of data has not yielded any concrete results because it's too complex for simple statistical methods. 3) Quantitative finance: from the failure of stat arb funds in 2007 to the blind trust in BSM in derivatives risk management software, mathematical models have beautiful assumptions that the world is smooth whereas the real world is not. Emergence of Hardware: Fab labs, 3D printers will enable delivery of real consumer products via CAD schema. Alternative energy research, medical robotics for remote surgical operations. Just look at the real trailblazers of Silicon Valley, SpaceX, Tesla and Google Driver-less cars. All hardward projects. Personally, I'd like to bone up EE, molecular biology lab protocols, CAD, MechE skills instead of RoR or Facebook APIz.

mgummelt 14y ago
I would assume driver-less cars are primarily a software achievement. Same with robotic surgery, but I'm not sure. Also, you can't compare two things by comparing the failures of one to the successes of the other.
michaelochurch 14y ago
I think AI has achieved quite a lot. Consider Google and Amazon, much less Watson. AI is like "magic". Whenever we figure out how to do something, it's no longer AI. It's statistical learning or computer vision. However, the problem of general AI turned out to be a lot harder than we thought it would be. Moravec's Paradox deserves mention here. Hardware and software will interact more in the 2010s to '30s. What will people design all this awesome hardware with? Software. Taking the other direction, I do agree that the hardware world is becoming more interesting to people who might, a generation ago, have thought of themselves as "pure software" people. In 1985, you could "do AI" and never go outside of a garbage-collected language (Lisp). In 2013, it's hard to do high-performance machine learning without having some knowledge of all the neat things people can do with hardware (e.g. GPU programming). Also, the existence of continuing education options like Coursera is making people more Y-shaped than T- or I-shaped. I agree with you that a lot of the "hot" work in the social media world (learning a bunch of damn APIs) isn't very hard, deep, or interesting. However, don't rule out software entirely.
- SatvikBeri 14y ago
  I understand T-shaped, but what are Y-shaped and I-shaped?
  - michaelochurch 14y ago
    I-shaped: depth with no breadth (highly specialized, no cross-disciplinary knowledge). Y-shaped: some breadth at intermediate levels. More depth in one's non-specialty areas than T-shaped. (The top of the T becomes a triangular funnel instead of a bar.)
    - SatvikBeri 14y ago
      Thanks! One thing I've noticed is that many of the most successful people are multi-specialists, i.e., they have two (or more) "specialties" that combine in a rare and valuable way. Examples include the rise of the "growth hacker" or "data scientist", who are typically programmers applying their craft towards marketing or statistics respectively. There's both a perception component and a real value component. If you're the best Marketer, other Marketers will frequently deny it, get jealous, or sabotage you. But if you're "the Marketer who also writes code", nobody feels threatened. You're competing in a different area, so people can support you and ask you for favors when they need help. In terms of real value, a junior programmer with strong business skills in a company with no other programmers can contribute millions of dollars of value through basic automation-I've actually done this. The same programmer will have a much lower dollar value when surrounded by better programmers, at least initially, because all the low-hanging fruit has been taken.
      - michaelochurch 14y ago
        Excellent observation. This relates to my observation that it's better to be an "X Who Programs" than a "J.A.P." (Just A Programmer) in most cases. Unfortunately, corporate HR systems are built on the idea that people need to be reduced to a single, specific role. Dimensionality is the rule, here. We suffer from it in hiring, where it's literally impossible to meet perfectly the requirement lists many companies put out (demanding 5 years of experience in a 3-year-old NoSQL database). This seems to be a part of why hiring is so hard on both sides. Yet, once people are working together, if people aren't careful, they forget dimensionality altogether and attempt to form a total order, and it's this process that leads to the insecurity and sabotage.

mgummelt 14y ago

I would assume driver-less cars are primarily a software achievement. Same with robotic surgery, but I'm not sure. Also, you can't compare two things by comparing the failures of one to the successes of the other.

michaelochurch 14y ago

I think AI has achieved quite a lot. Consider Google and Amazon, much less Watson. AI is like "magic". Whenever we figure out how to do something, it's no longer AI. It's statistical learning or computer vision. However, the problem of general AI turned out to be a lot harder than we thought it would be. Moravec's Paradox deserves mention here. Hardware and software will interact more in the 2010s to '30s. What will people design all this awesome hardware with? Software. Taking the other direction, I do agree that the hardware world is becoming more interesting to people who might, a generation ago, have thought of themselves as "pure software" people. In 1985, you could "do AI" and never go outside of a garbage-collected language (Lisp). In 2013, it's hard to do high-performance machine learning without having some knowledge of all the neat things people can do with hardware (e.g. GPU programming). Also, the existence of continuing education options like Coursera is making people more Y-shaped than T- or I-shaped. I agree with you that a lot of the "hot" work in the social media world (learning a bunch of damn APIs) isn't very hard, deep, or interesting. However, don't rule out software entirely.

SatvikBeri 14y ago
I understand T-shaped, but what are Y-shaped and I-shaped?
- michaelochurch 14y ago
  I-shaped: depth with no breadth (highly specialized, no cross-disciplinary knowledge). Y-shaped: some breadth at intermediate levels. More depth in one's non-specialty areas than T-shaped. (The top of the T becomes a triangular funnel instead of a bar.)
  - SatvikBeri 14y ago
    Thanks! One thing I've noticed is that many of the most successful people are multi-specialists, i.e., they have two (or more) "specialties" that combine in a rare and valuable way. Examples include the rise of the "growth hacker" or "data scientist", who are typically programmers applying their craft towards marketing or statistics respectively. There's both a perception component and a real value component. If you're the best Marketer, other Marketers will frequently deny it, get jealous, or sabotage you. But if you're "the Marketer who also writes code", nobody feels threatened. You're competing in a different area, so people can support you and ask you for favors when they need help. In terms of real value, a junior programmer with strong business skills in a company with no other programmers can contribute millions of dollars of value through basic automation-I've actually done this. The same programmer will have a much lower dollar value when surrounded by better programmers, at least initially, because all the low-hanging fruit has been taken.
    - michaelochurch 14y ago
      Excellent observation. This relates to my observation that it's better to be an "X Who Programs" than a "J.A.P." (Just A Programmer) in most cases. Unfortunately, corporate HR systems are built on the idea that people need to be reduced to a single, specific role. Dimensionality is the rule, here. We suffer from it in hiring, where it's literally impossible to meet perfectly the requirement lists many companies put out (demanding 5 years of experience in a 3-year-old NoSQL database). This seems to be a part of why hiring is so hard on both sides. Yet, once people are working together, if people aren't careful, they forget dimensionality altogether and attempt to form a total order, and it's this process that leads to the insecurity and sabotage.