What do tech pioneers think about the AI revolution? - BBC World Service

Key Concepts:

1. Introduction and Panel Overview:

Caroline Steel hosts "The Engineers" at Imperial College London, focusing on the AI revolution.
Panelists:
- Paolo Pirjanian (Embodied): Builds emotionally intelligent robots for child development.
- David Silver (Google DeepMind): Led the team that defeated the world's best Go player and is working on AGI.
- Regina Barzilay (MIT): Pioneered AI for early breast cancer detection and antibiotic discovery.
The program highlights AI's impact, from strategy games to medicine.

2. Regina Barzilay's Shift to Oncology:

Barzilay, a natural language processing and AI expert, was diagnosed with breast cancer in 2014.
She observed a lack of AI in cancer treatment (diagnostics, treatment, post-treatment).
Motivated by her experience and the suffering of other patients, she sought to apply AI to healthcare.
She collaborated with Dr. Connie Lehman at Massachusetts General Hospital (MGH) to focus on early cancer detection.
Example: Barzilay's personal experience with breast cancer treatment motivated her to apply AI to healthcare.

3. David Silver and Reinforcement Learning:

Silver's initial disappointment with AI in computer games led him to reinforcement learning.
Reinforcement learning: A system learns from experience through trial and error, similar to how humans and animals learn.
Analogy: Learning not to touch fire due to the negative consequence of pain.
Machines are rewarded with numerical values (positive for good actions, negative for bad actions).
The "reward" signal drives the learning process.

4. Paolo Pirjanian and Emotionally Intelligent Robots:

Pirjanian's experiences of alienation in foreign countries inspired him to create a robot companion.
He noted the growing prevalence of autism (1 in 30 children) and the need for social skills training.
Embodied's robots are designed to understand human emotions and build relationships with children.
The robots help children practice social skills like eye contact, turn-taking, and joint attention.
Example: Embodied's robots are designed to help children with autism spectrum disorder develop social skills.

5. AI's Capabilities in Cancer Understanding (Regina Barzilay):

AI can handle uncertainty in cancer diagnosis and treatment, which humans struggle with.
AI can analyze vast amounts of data to improve predictions and reduce errors.
Example: AI detected early-stage cancer in Barzilay's mammograms two years before it was visible to radiologists.
AI can remove the "guessing" element from diagnosis and treatment decisions.

6. AI and the Game of Go (David Silver):

Go is more challenging for computers than chess due to its complexity and reliance on intuition.
Human Go players rely on intuition and creativity, which are difficult to replicate in machines.
Chess can be solved with tactical look-ahead, but Go requires imagining the game's progression hundreds of moves in advance.
AlphaGo required major breakthroughs to achieve human-level intuition.

7. Emotional Bonds with Robots (Paolo Pirjanian):

Robots are not meant to replace human contact but to serve as "training wheels" for social skills.
Humans are wired to form bonds, even with inanimate objects.
Robots with eyes, eye contact, smiles, and emotional expression can quickly form bonds with children.
Children may open up to robots more easily than to therapists or parents.

8. AI for Antibiotic Discovery (Regina Barzilay):

Pharmaceutical companies are less active in antibiotic development due to economic factors.
AI can help discover new molecules that are effective against drug-resistant bacteria and non-toxic to humans.
MIT's AI model screened thousands of molecules and predicted their effectiveness.
The AI discovered a molecule that kills bacteria using a different mechanism of action, making it effective against multiple species.
Process: AI analyzes molecular structures to predict their antibacterial effects.

9. Artificial General Intelligence (AGI) (David Silver):

AGI aims to create AI systems that can approach any number of problems with intelligence, similar to humans.
AGI is a spectrum and requires further breakthroughs to reach human-level intelligence.
AGI systems should be able to learn diverse skills and specialize in different areas.

10. Challenges in AI Adoption in Medicine (Regina Barzilay):

Technology is mature, but translation into patient care is lacking.
Regulations in Europe, the UK, and the US are constantly changing.
Lack of clear billing practices for AI use in medicine.
Example: Doctors in the US may lose money by using AI because it reduces the time spent with patients.

11. Future of Social-Emotional AI (Paolo Pirjanian):

Social-emotional AI can help humanity become its best.
Early intervention with children on the autism spectrum can improve their integration into society.
AI companions can address social isolation and loneliness in aging populations.
Robots can provide assistive care, including cooking and walking, to promote independent living with dignity.
Assistive care robots are expected within the next decade.

12. AI and Creative Tools (David Silver):

AI will likely provide powerful tools to human authors rather than taking over human culture.
Example: Lyria, a music authoring system, speeds up the songwriting process.
AI and humans can work together to create better novels and other forms of media.

13. Regulation of AI (Panel Discussion):

Simon asks about the British government's approach to AI regulation.
David Silver: Regulation is important but should be tailored to different areas (e.g., medicine vs. chatbots).
Regina Barzilay: Concerns about hindering progress in treating diseases due to slow regulation.
Paolo Pirjanian: Regulation could create a strategic disadvantage if adversaries don't follow the same rules; international agreement is needed.

14. AI and Sports Performance (David Silver):

AI is being used to improve sports performance.
Example: Google DeepMind is collaborating with Liverpool Football Club to improve their tactics.

15. AI and Human Learning (Panel Discussion):

A young audience member asks if AI will hinder human learning.
David Silver: AGI could be a personal friend, assistant, and teacher to enhance learning.
Regina Barzilay: AI can remove pressure from tasks like writing, allowing focus on ideas.
Paolo Pirjanian: AI can make individuals more prolific by automating tasks like coding.

16. Conclusion: