The peak of A.I. ecosystem and what’s next?

This week I want to comment on Tom Rikert’s excellent article “AI Has Peaked — What’s Next?” I’m seeing a similar thing in the A.I. ecosystem, which is, many of the products launched, for all their “intelligence” show limited marginal value to users. This makes adoption more difficult. Tom’s example of Tools vs Force Multipliers vs Automation is a powerful framework to use when thinking about what your A.I. product does and whether adoption will be easy or hard.

But the bigger problem is, the state of the A.I. industry is still such that given a proprietary data set, it is often hard to predict both the quality of outcome you can deliver, and how much you can improve upon it with more data.

To expand on the first point, many startups are trying something they see in an academic research paper, and the gap between academic A.I., with it’s well controlled technical environments, super clean data sets, and lack of need to deal with the messiness of real user behavior, means that many of the amazing breakthroughs you read about are still years away from being incorporated into real production systems that have real daily users. It’s common to see a team that thinks they can get some A.I. driven metric to a high accuracy level because they read the results of a research paper that did so, but when you lift the hood, it’s unclear if that will ever really happen in production.

And secondly, even if they can, how fast will the algorithm improve with more data? I wish there was a Moore’s Law type concept that predicts the future performance of a model given more data (e.g. double the data and get X% improvement in your output) but there isn’t (so far). This means that when you have a model that is 70% accurate, and you augment with humans in the loop for the other 30%, you often have no idea whether the model will get to 95% accuracy in one year, five years, or never.

These things make starting A.I. companies, and investing in A.I. companies, difficult. But given the potential returns, for the ones who stumble on to the right areas and solve the right problems, I still believe it is highly worth it. Tom’s points from the article are good to think about as you navigate the A.I. landscape, and I still believe as the tools evolve we will see more useful frameworks develop for thinking through these opportunities.


In other news….

Facebook’s Instagram is partnering with BigCommerce to make shopping links more widely available. The partnership allows merchants using the BigCommerce sales platform to tag products on Instagram posts. Tapping on the tag reveals product information and a link to buy it on the merchant’s website. Merchants will also have access to data from Instagram, including the number of people who viewed the information or clicked to the product page.

Microsoft acquires struggling social VR startup AltspaceVR, terms undisclosed; the company enables users to meet in VR space and participate in various activities; had contract with NBC to broadcast the presidential debates; faced tough competition from services such as Facebook Spaces

The era of the robots — The International Federation of Robotics revealed that global sales of industrial robots rose by 16 percent in 2016. Has the era of the robots arrived and should we fear significant job losses?

https://www.reuters.com/article/us-robotics-industrial-statistics/electronics-industry-drives-2016-industrial-robot-sales-ifr-idUSKCN1C22OR

God is a bot — Anthony Levandowski, a multi-millionaire Uber engineer, has founded a religious organisation called Way of the Future, with the sole objective to “develop and promote the realisation of a Godhead based on Artificial Intelligence.”

Self-healing databases — In case you do not want to rely on your engineers to fix your databases, teach AI to do the job. Oracle has announced a database that uses AI to fix itself.

Vuforia, a Qualcomm subsidiary now owned by PTC, launched a new version of its AR platform and an app called Chalk. Chalk is a videoconferencing app that allows digital graphics to be “drawn” on the display in real time using AR. The app is free, but Vuforia hopes to sell the underlying AR technology to businesses. The new version of the company’s AR platform, Vuforia 7, also enables developers to write a single AR app that will work on most devices, including Apple and Android.

Virtual Reality and Augmented Reality device sales are forecasted to hit $11.9 billion in 2021. According to a report by CCS Insights, 99 million devices will be sold in 2021. The analysts also predict that 16 million devices will be sold in 2017, a 47% increase over the previous year. Smartphone VR and AR devices will account for more than 13 million units, and most of the 2017 $1.6 billion market value will be derived from dedicated VR hardware like the Oculus Rift. Microsoft’s Mixed Reality headsets are expected to boost the market in Q4 2017 and in future years.

The U.K. is looking for solutions to destroy contraband drones headed to prisons. The Ministry of Justice, the entity which manages the country’s prisons, has launched a competition looking for ideas to stop contraband from dropping via drone. The MOJ wants to use technology as a solution to this problem and is dedicating $1.2 million in a fund for potential solutions. The department wants solutions for detecting contraband items, like mobile phones, and for capturing and destruction of drones headed into prisons. Interested applicants may apply for a $66,241 grant to develop their designs

Last year, F-Secure and the University of Helsinki launched an online course series called Cyber Security Base. It’s a free, massive open online course designed to give people the knowledge and skills they need to begin a career in cyber security. Nearly 5000 people took part in the course assignments last year. The course will start on October 31.

An Ireland researcher thinks there might be a chance other life forms could be taking a peek at Earth — possibly without us knowing.

The chances could be slim, though, since most of our galaxy neighbors likely would — quite literally — need the planets to be aligned to see here. We have a similar probability of being nosey enough to see them, however.

While we know there are other planets, and even some that could offer life, what’s unknown is the scope of their technology. If their tools are far more advanced, they might already know a lot about us. On the other hand, they could be in a primitive state and not have much info on us.

Elon Musk last week announced ambitious plans that include building a base camp on the moon and traveling anywhere around the Earth in about 30 minutes. The SpaceX chief executive said passengers could get from New York to Shanghai in as little as 39 minutes traveling at speeds of at least 16,000 mph. His announcement was part of the International Astronautical Congress in Adelaide, Australia. Musk plans two private trips around the moon next year. “If you want to get the public really fired up, I think we’ve got to have a base on the moon. That’d be pretty cool. And then going beyond there and getting people to Mars. That’s the continuance of the dream of Apollo that I think people are really looking for,” Musk said.

“Is A.I. Riding a One Trick Pony?” It looks at the current wave of A.I., mostly driven by decades old innovations in neural networks, and questions whether we have another advance waiting in the wings to push the industry forward when deep learning peaks. I’m a lot more optimistic, because I think the reinforcement learning stuff that is starting to become more prevalent will take neural networks to a whole new level before we have to worry about the stagnation that comes from marginal improvements. But it’s always good to read into other (particularly more skeptical) points of view.

Robots could destabilise world through war and unemployment, says UN. Guardian — There is a new headquarters to monitor AI developments being established at The Hague. The goal is to protect the populace against the possible threats of mass unemployment or rogue use of autonomous weapons.

Gains from AI could mean humans live for leisure some day. Open Democracy — While a lot of people are focused on the fallout from AI — unemployment and weapons like the group above — Miles Brundage, a specialist in Artificial Intelligence and Policy and research fellow at the Oxford Future of Humanity Institute, imagines a world where automation allows humans to focus on creative and leisure activities.

Companies need to stop A.I. washing their products. — A great look at the overuse of saying things are “A.I.” without saying what that means for your customers. This trend is why the market is so confused.


Thanks for reading! :) If you enjoyed this article, hit that clap button below 👏 Would mean a lot to me and it helps other people see the story.

Say Hello On

Instagram | Twitter | Facebook | SnapchatLinkedIn

One clap, two clap, three clap, forty?

By clapping more or less, you can signal to us which stories really stand out.