Autonomous Systems: From Fiction to Reality

Daniel Raviv, Ph.D.
Professor, Department of Computer and Electrical Engineering and Computer Science
College of Engineering and Computer Science

How soon will self driving be completely reliable and like human reactions and decisions?

Depends on the driving environment. For "normal" highway and city driving it is already there. Take a look at Mobileye, Google and Tesla.

If laws allowed it, do you think autonomous cars could work on our road today?

Yes. Autonomous cars have driven miilion of miles already, and very successfully.

Isn't one of the most complicated problems with autonomous vehicles when a human tries to override the system? If you have not already answered this question, maybe you have an opinion?

It can be designed in a way that human can easily take over.

What are some differences from working with cutting edge technology in the US and other countries?

I have a limited experience in working in other countries. In the US we tend to "invent technology," take risks, and learn from failure.

Given the scope of racial inequities endured by black Americans, and disadvantaged people around the world, the implementation of A.I. presents as a viable soultion to eliminate race-based bias in socio-economic, education and judicial sectors. What are the plans, if any, to deploy A.I. in address of the above?

For a more general answer by experts, please refer (for example) to "Talking to Strangers" book by M. Gladwell.

IBM Watson playing Jeopardy was a scam. Watson got the prompts in text form. No speech-to-text, and much faster than the host could read them aloud to the humans. In that game, a 0.5 second advantage is huge.

Machines have advantages (reaction time, search time, processing time). "The NO speech to text" is new to me - it can indeed cause meaningful disadvantages.

The productivity boost resulting from the Industrial Revolution is a LOT steeper than the curve you show.

Sure. The curve is meant to be qualitative (with no scale).

Does the machine learning become as effective as it does because it has a lack of human emotion?

Machines can handle huge amount of data (+fast and cheap). And this is only the beginning.

I fly drones and have not been able to fly in remote areas that don't have wifi or strong wifi at least. What technology needs to be in place for autonomous flight and operations to take place?

WiFi is only one well known technolgy. There are many alternatives (e.g., GPS), usually expensive. If you try to fly them long distances, there is a problem.

Do any of your vehicles employ vehicles communicating with each other as they approach an intersection to optimize collision avoidance?

My research focuses on local perception of the environment. So the answer to your question is no. Look at V2V-related research.

Thirty years ago I learned LISP to be ready for AI. What languages are used these days?

Python has become very popular.

Do you project that quantum computing will accelerate AI algorithm development, and in what timeframe?

Quantum computing will change "everything" but we are not there yet. There is a great book, "The Code," by Simon Singh - the last chapter is about Quantum Computing.

Are quantum computers useful in AI Learning?

We are not there yet, but surely yes.

Do you have assitanceships in your department for B.S. M.S students?

Please contact the CEECS department at FAU.

How can we make sure in the future is our students taking classes and taking exams, not machine taking classes ?

Excellent question! But I do not have an answer. I feel that the education system will become very very different.

How can students prepare in highschool themselves to enter AI discipline?

Focus on STEM, and in particular on Math and Science.

How long until machines will be able to change their own code and (evolve) just as good or possibly better than humans can?

It is happening already. Machines improve their learning with experience.

When do you predict this technology will be available to public? When do you think driverless cars will become mainstream?

In 1993 I predicted that we all "see" autonomous cars in 2002. Now it has become reality (US and Germany are among the leaders).

How will autonomus vehicles and their vision systems be able to adapt to objects in extreme weather conditions in various parts of world? For example in a blizzard conditions.

The learning process is on going. There are many hundreds(!) of companies that work on this and other related problems.