Artificial Intelligence Bulletin – Inflated Expectations

Posted By on June 8, 2016

Screen Shot 2016-06-06 at 8.51.20 AM

Each week on a Wednesday Nikolas Badminton, Futurist highlights the top stories from the past week relating to the incredible rise of artificial intelligence and its application in society, the workplace, in cities, and in our lives.

In Artificial Intelligence Bulletin – Inflated Expectations we see posts on robots stealing jobs, the age of brain emulation, doctors going out of business, and Amazon stepping back.

Federico Pistono, author “Robots Will Steal Your Job, But That’s OK” AMA

There was a great Reddit AMA this week – [AMA] I am Federico Pistono, author “Robots Will Steal Your Job, But That’s OK” and “How to Create a Malevolent Artificial Intelligence” with prof. Yampolskiy. Ask me Anything!

Especially the first question

Chispy

Most of the progress we’ve seen in machine learning in recent years haven’t had much of a noticeable effect on how the average person lives their lives. Could this change over the next few years? What are some promising applications for AI and how disruptive do you think it’s going to be over the next 5-10 years?

federicopistono

Great question!

First, I’m not sure I completely share your assumption. It is true that at first sight one might not notice what impact recent advancements on machine learning have had on their lives. And if you ask what were some revolutionary technologies, many experts would say the airplane, refrigeration, sanitation, and other very tangible, physical things. Changes brought by better machine learning algorithms are not as immediately noticeable, but could be just as important.

As I see it, there are three aspects at play.

(1) We’re bad at managing expectation. We’re social animals, influenced by the media, by what our friends say and think, and neither are (usually) good predictors of technological progress and its impact on society. We overestimate in the short-term what a new tech can do (we have a feeling it will change the world in 5 years or so), and we underestimate what it can actually do in the medium or long term.https://upload.wikimedia.org/wikipedia/commons/thumb/b/bf/Hype-Cycle-General.png/1024px-Hype-Cycle-General.png The reality is that things that time, and new technologies have an incubation period of anything between 10 to 30 years, before they’re ubiquitous and cheap. While it’s true that this cycle has been shortening, there are still some constraints that prevent it from going below the 5-10 year limit.

Watson beat the best *Jeopardy! players? OMG it’s gonna change everything!* — after 5 years, we see some applications, but nothing revolutionary and nothing at large scale yet. Google unveils the first autonomous car, now this is going to change everything!!! — five years in, fully autonomous cars are still not available at your local car dealer.

And so we lose faith in what a specific technology can do, not realizing that in a few more years we’re going to see the real impact it will have.

(2) We’re bad at quantifying impact. Keeping in mind (1), we’re also bad at recognizing what has already happened and how many people have been affected by it, because once something’s out there for everyone, we take it for granted. You might not think much of it, but how many lives were improved by better prediction algorithms in the distribution of energy in the grid? How many more by better detection of cancerous cells and other anomalies? Improved weather forecast impacts the lives of billions of people, many of whom rely on farming and simple commerce to survive, yet we don’t hear of it in the media.

On the flip side, face detection, data classification and metadata analysis by security agencies have had a very different yet profound impact on our lives. We’re part of a global surveillance state, whether we like it or not, and we’re just beginning to have the much needed conversation about it.

In today’s hyper-connected world, when you think of impact, think more of systems, and less of consumer products.

(3) AFAIK, there haven’t been major breakthroughs in basic AI research. While we had impressive improvements in applied research for narrow AI – machine learning, deep learning, neural networks, switch to GPUs, more powerful machines. etc. – basic research seems to have had very few breakthroughs. This is true, and it’s why I believe that true AGI is still pretty far.

As for your specific question, I think in the next 5-10 years we’re going to see the fruits of what we planted 5-10 years ago: human-level speech recognition, much better prediction systems, medical diagnoses, better organizations systems, widespread self-driving cars, etc.

We’ll hear a lot of hype about the “new things at the door” (personalized medicine, rewriting your genome, AGI, etc.) and will follow the same patter as I described in (1), thus perpetuating the hype-cycle of inflated expectations.

Read more at Reddit

The Age of Em – with Robin Hanson

Part 1 in interview with Robin Hanson on The Age of Em. We cover the nature of ems and their social systems, how they compare to AI and whether they might be the only form of AGI for some time in the future, approaches to prediction (generally & specific to the book), and particular concrete predictions about the impacts of emulations on our civilization tomorrow and into the far future

We also discusses Scott Alexander’s review of The Age of Em (at Slate Stellar Codex): http://slatestarcodex.com/2016/05/28/…
Robin Hanson’s blog response: http://www.overcomingbias.com/2016/05…

It’s man versus robot in the battle of the doctors: World’s first ‘artificial intelligence’ medic set to be pitted against the real thing in landmark experiment for medicine

The world’s first ‘artificial intelligence’ doctor will be pitted against the real thing this week, in a head-to-head contest that could mark a turning point in medicine.

British start-up firm Babylon Health will test its programme, called Check, against a doctor and nurse in a competition to see which can deal most quickly and accurately with a range of common health problems.

The smartphone app has been designed to act like a triage nurse, asking a series of questions to advise users whether their problem is nothing to worry about, something they should consult their GP about, or a matter that requires calling 999.

Read more at The Daily Mail

Why artificial intelligence probably isn’t Amazon’s next coup

Amazon’s Jeff Bezos has talked big game when it comes to artificial intelligence. That said, it’s shipping, business supplies and groceries that are likely to be Amazon‘s next big business, one technology analyst told CNBC’s “Squawk Alley” recently.

Read more at CNBC


Like the story? Post comment using disqus.