Lifeboat Foundation

There was an historic first today as we welcomed an actual, real-life robot onto the sofa for a chat. 🤖 Sophia not only drew pictures of Holly and Phillip, she also summoned the memory of Gordon the Gopher! 😂😂.

A schism lies at the heart of the field of artificial intelligence. Since its inception, the field has been defined by an intellectual tug-of-war between two opposing philosophies: connectionism and symbolism. These two camps have deeply divergent visions as to how to “solve” intelligence, with differing research agendas and sometimes bitter relations.

Today, connectionism dominates the world of AI. The emergence of deep learning, which is a quintessentially connectionist technique, has driven the worldwide explosion in AI activity and funding over the past decade. Deep learning’s recent accomplishments have been nothing short of astonishing. Yet as deep learning spreads, its limitations are becoming increasingly evident.

If AI is to reach its full potential going forward, a reconciliation between connectionism and symbolism is essential. Thankfully, in both academic and commercial settings, research efforts that fuse these two traditionally opposed approaches are beginning to emerge. Such synthesis may well represent the future of artificial intelligence.

Just fire everybody?

Synthetic protocells can be made to move toward and away from chemical signals, an important step for the development of new drug-delivery systems that could target specific locations in the body. By coating the surface of the protocells with enzymes—proteins that catalyze chemical reactions—a team of researchers at Penn State was able to control the direction of the protocell’s movement in a chemical gradient in a microfluidic device. A paper describing the research appears November 18, 2019 in the journal Nature Nanotechnology.

“The futuristic vision is to have drugs delivered by tiny ‘bots’ that can transport the drug to the specific location where it is needed,” said Ayusman Sen, the Verne M. Willaman Professor of Chemistry at Penn State and the leader of the research team. “Currently, if you take an antibiotic for an infection in your leg, it diffuses throughout your entire body. So, you have to take a higher dose in order to get enough of the antibiotic to your leg where it is needed. If we can control the directional movement of a drug-delivery system, we not only reduce the amount of the drug required but also can increase its speed of delivery.”

One way to address controlling direction is for the drug-delivery system to recognize and move towards specific chemical signals emanating from the infection site, a phenomenon called chemotaxis. Many organisms use chemotaxis as a survival strategy, to find food or escape toxins. Previous work had shown that enzymes undergo chemotactic movement because the reactions they catalyze produce energy that can be harnessed. However, most of that work had focused on positive chemotaxis, movement towards a chemical. Until now, little work had been done looking at negative chemotaxis. “Tunable” chemotaxis—the ability to control movement direction, towards and away from different chemical signals—had never been demonstrated.

“The Hyperloop exists,” says Josh Giegel, co-founder and chief technology officer of Hyperloop One, “because of the rapid acceleration of power electronics, computational modeling, material sciences, and 3D printing.”

Thanks to these convergences, there are now ten major Hyperloop One projects—in various stages of development—spread across the globe. Chicago to DC in 35 minutes. Pune to Mumbai in 25 minutes. According to Giegel, “Hyperloop is targeting certification in 2023. By 2025, the company plans to have multiple projects under construction and running initial passenger testing.”

So think about this timetable: Autonomous car rollouts by 2020. Hyperloop certification and aerial ridesharing by 2023. By 2025—going on vacation might have a totally different meaning. Going to work most definitely will.

Lawyers and doctors are typically paid more than manual laborers because of the relative shorter supply of lawyers and doctors, which is in part due to the number of years of training required to enter those professions and the corresponding value society attributes to those skills. But what will happen to their wages once the market is faced with an abundance of skilled labor? If anyone is able to upload legal or medical know-how to their brain and know just as much as the professionals in those fields, why pay a professional a higher wage?

Of course, certain skills, such as strategic judgment and contextual understanding, may be difficult, if not impossible, to digitize. But even the games of chess and Go, both complex games that require strategic decision-making and foresight, have now been conquered by AIs that taught themselves how to play—and beat—some of the best human players.

The technology’s potential for emancipation and human advancement is immense. But we—entrepreneurs, researchers, professionals, policymakers, and industry—must not lose sight of the social risks.

Maybe interesting for this group too.

Visit http://TED.com to get our entire library of TED Talks, transcripts, translations, personalized Talk recommendations and more.

Continue reading “The danger of AI is weirder than you think | Janelle Shane” »

Chun Yuan Chiang of IHDpay Group says artificial intelligence cannot completely replace the “high-touch” nature of medical care. However, technology can be helpful in diagnosis or in situations where patients have long, complicated medical histories, he says. Chiang was speaking on a panel with Jai Verma of Cigna International and Dai Ying of GE Healthcare.

Most humans can learn how to complete a given task by observing another person perform it just once. Robots that are programmed to learn by imitating humans, however, typically need to be trained on a series of human demonstrations before they can effectively reproduce the desired behavior.

Researchers were recently able to teach robots to execute new tasks by having them observe a single human demonstration, using meta-learning approaches. However, these learning techniques typically require real-world data that can be expensive and difficult to collect.

Continue reading “An imitation learning approach to train robots without the need for real human demonstrations” »