What after AI?

Now a days most of the companies are working on AI. I feel it will reach saturation in couple of years and most of our jobs will be done by AI What do you think people/companies will start working on after that?

New
Oich87 Jul 31, 2017

Good question.

Amazon homo Jul 31, 2017

Homo deus

New
staywke Jul 31, 2017

Universal Basic Income

Amazon homo Jul 31, 2017

New world order - dramatic shift in public policy. Who knows may be mass migration to mars!

Google ab12 Jul 31, 2017

killing a.i. ☺

Google ab12 Jul 31, 2017

checkout this movie from the last century called terminator...

Amazon Putin Jul 31, 2017

Good one comrade

Microsoft C. Kramer Jul 31, 2017

Nanotech.

Amazon RubbaDuck Jul 31, 2017

I think we are far from saturation. We've figured out autodifferentiation APIs for like 6 years. There's multiple hardware manufacturers and clouds that are pursuing ASICs for cheap/fast tensor-graph computation. Parameter servers and hyper parameter optimization are sort of solved. Differential neural computers show progress in memory augmented networks. I think the next step is determining the RPC mechanism for single responsibility neural networks. We can't keep scaling the unit of training and we can't repurpose existing RPC mechanisms to grow natural understanding. We need a vector based, dare I say, dynamic RPC to get us through the next phase of natural understanding. This will allow us to train automata and scale teams to impact. After that we need neural nets to understand the control plane of these systems. I see it as a slog that will result in highly visible failures, the same way that the query execution planners of yesteryear caused service outages and high latency. Next comes advice systems. These advice systems augment our personal understanding of the digital world. Beyond that I don't know.

Google 😳😶😳😂3 Jul 31, 2017

Beyond that, the AI will decide.

Amazon RubbaDuck Jul 31, 2017

I think that the advice systems will allow humans to scale the impact of changes to the control plane of their data layers. There's a tension when you involve "actuation" in the process though. If the actuation is highly impactful then the probability of the correct interpretation becomes more scrutinized.

Uber GMie27 Jul 31, 2017

The bottleneck will move from feature engineering to data collection. Data Scientists will evolve to AI trainers.

Google 😳😶😳😂3 Jul 31, 2017

"machine teacher" job openings?

Uber GMie27 Jul 31, 2017

Yes, defensive driving class for self-driving cars.

Slalom Consulting Samms Jul 31, 2017

Self driving cars,smarter home automation, solar, agritech imho