NuclearPhynance Archive

Self-driving cars are the wildcard. It's the only tech on the horizon that has a realistic chance of having a meaningful economic impact. For all the hype surrounding apps and mobile and social networking, nothing has actually had any real affect on total factor productivity since email and spreadsheets.



Information technology did substantially grow productivity in the 90s, at almost the same rate as the first and second industrial revolutions. Unfortunately it proved much shorter lived than its antecedents. After the initial wave of wide-scale PC adoption, we've returned to pre-computer productivity trends. The info-tech industry is basically just making new toys rather than new tools at this point.



Self-driving cars could really buck this trend. In the US at least 5% of the labor force could be put out of a job. Not just truck drivers, deliverymen and bus drivers. It also means less auto manufacturing (most people can dispense with owning a full time car), fewer body shops and mechanics (less accidents), fewer ER docs/nurses (less car accidents), a much smaller car insurance industry, less road construction (less traffic), and even less housing construction (people can just move further away, and sleep during their commute).



That being said, I've lately heard a lot of skepticism from insiders about whether the tech is anywhere on the horizon. It might be a lot like how speech recognition seemed deceptively close in the 1980s. It was easy to get the accuracy from 0% to 95%. The threshold for usability is 99%, so it seemed like success was imminent. But the last mile ended up being way harder than prior work.

I second jslade on the AI winter. I've been part of a couple highly disfunctional big data/ai/machine learning projects at big companies and have lots of connections that tell me familiar stories.



It's not that the ML part is not working at all...it's just that there are crazy to fradulent claims on power point presentations and hence "C" types have absolutely crazy expectations on what ML + data can deliver.



And hence the projects most of the time do not deliver to the point that they are realy adding substantial comercial value to the existing solutions.



"C" types are very binary...and hence I think they will flip the switch soon and label everything that has to do with machine learning as evil and academic nonsense.



Disclaimer: this is obviously not true for many things in robotics and control theory related applications of AI that work incredibly well.

> Unfortunately it proved much shorter lived than its antecedents.



@ExpressoLover It is not still going? Automation everywhere is increasing (for example supermarket self service) and computing power has only got really cheap in the last 5 years

AI winter? Who, with a minimum experience of the topic, would have bought in the ML/AI hype in the first place anyway?

Management of large companies? VC type idiots?



From what I read from you in other threads you are quite independent..which is great for you.



If you are not an independent researcher , for example a "consultant" or an academic you need someone who pays you for doing fun stuff. If that moneyflow stops a lot of stupid stuff stops for the right reasons. But some good stuff is also canceled.

@Avenger

> I think the medical sector has a lot of potential growth. But im 19, what do i know



The space of possibilities is limited by imagination of those who are 19...

...and don't know about the crazy regulation hurdles in healthcare

To the point )))

goldorak: "Who, with a minimum experience of the topic, would have bought in the ML/AI hype in the first place anyway?"



I trace it back to the "DeepMind" acquisition by Google. Basically a textbook reinforcement learning example (playing atari games) that looked kewul, especially to over-wealthy Gen-X people in bizdev who played atari games as children. Once that happened, VCs had to be in on the game. Once VCs got in on the game, the hype bandwagon took on a life of its own.



I think it's fair that ML is more accessible to the masses at least, but the whole AI apocalypse thing is just a joke. ML replaces large teams of statisticians, not really anything else.

@jslade, I heard that Facebook is building a block-chain based AI system and want to recruit people like you!