Steve Jurvetson: Forecasting the Future of Technology [Entire Talk]

TL;DR

Technology forecasting works best when predictions are treated with humility, weak ideas are tested and discarded, and persistent capability trends are separated from uncertain products or companies. The panel identifies machine learning, biology, robotics, and intelligent software-centered systems as major areas of progress, while emphasizing passion and long-term foundational research when choosing problems.

Transcript

So I get to have the fun of asking the questions to this really great panel with three, I'd like to point out, really different points of view. Steve's venture capital funding, cool new ideas. Astro's running probably one of the most inventive labs, but I think is it fair to say managing a lot these days. Maybe still there inventing. And Christina,... Read More

Key Insights

• Forecasting technology is inherently uncertain because even experienced investors, inventors, and researchers can misjudge consumer behavior, company evolution, and the speed of scientific progress. Reviewing failed predictions provides a practical source of humility when making new forecasts.
• Company evolution is difficult to anticipate because businesses can expand, pivot, or adapt beyond their original products. Steve Jurvetson expected Amazon to become obsolete and Netflix to fail, but acknowledged that he did not foresee how those companies would navigate broader industry transitions.
• The search for valuable innovations requires accepting that most ideas will be wrong. Astro Teller describes progress as clawing through piles of ideas to find rare gems, using his mistaken rejection of camera phones as a reminder against excessive certainty.
• Artificial intelligence capabilities can progress despite periods of public pessimism. Teller observed a predictable exponential increase in AI abilities during an AI winter, followed by continuing benefits in machine translation, speech-to-text systems, computer vision, robotics, and related areas.
• Machine learning is turning industrial sectors into software-centered businesses. Jurvetson identifies rockets, drones, robots, and satellites as examples where software and repurposed off-the-shelf hardware can transform industries that previously appeared to be unattractive industrial businesses.
• Robots are systems that combine sensing, computation, and actuation, not merely machines shaped like walking people. Teller applies this broader definition to stratospheric balloons, airborne wind turbines, self-driving cars, and self-flying package-delivery vehicles.
• Biology can inform the engineering of complex intelligent systems. Jurvetson suggests that biological processes provide both an existence proof and metaphorical inspiration for iterating on algorithms, accumulating intelligence, and increasing complexity over time.
• Academic research should address long-term foundational problems rather than focusing only on outcomes two to five years away. Christina Smolke recommends looking 10 to 20 years ahead, choosing transformative questions, and pursuing work that merits a researcher’s full commitment.

Questions & Answers

Q: How can technology leaders improve future forecasts?

Technology leaders can improve forecasts by treating confidence with humility, examining previous mistakes, and distinguishing persistent capability trends from uncertain product outcomes. The panel’s examples show that consumer behavior, company pivots, and industry transitions are especially difficult to predict. A productive forecasting process considers many ideas, expects most to be wrong, and searches systematically for the few possibilities with genuine promise.

Q: Why are successful technology companies hard to predict?

Successful technology companies are hard to predict because they may evolve beyond the products and business models that initially define them. Steve Jurvetson expected Amazon to become obsolete through distributed, federated, personalized systems and thought Netflix would fail as internet video replaced mailed DVDs. His mistake was not anticipating how those companies could expand, pivot, or grow during their respective industry transitions.

Q: What does Astro Teller’s camera-phone mistake teach?

Astro Teller’s camera-phone mistake demonstrates that strong intuition can be unreliable when predicting new consumer behavior. He was certain that people would not take pictures with phones or become obsessed with sending those pictures to friends. He uses that memory as a humility check and as evidence for his view that nearly everyone’s ideas are wrong most of the time.

Q: Why did artificial intelligence progress after an AI winter?

Artificial intelligence continued progressing because public pessimism did not alter the underlying trajectory of its capabilities. Teller describes AI as following a predictable exponential increase even when people declared the field dead or incapable of meaningful achievement. He points to steady benefits in machine translation, speech-to-text technology, computer vision, robotics, and several other areas as evidence of that continuing development.

Q: How is machine learning transforming industrial sectors?

Machine learning is transforming industrial sectors by making products and businesses increasingly centered on software and intelligence. Jurvetson cites rockets, drones, robots, and satellites, industries that once looked like poor industrial businesses. These systems can combine machine-learning methods with off-the-shelf hardware repurposed from products such as cellphones, allowing new capabilities to emerge across a broad range of fields.

Q: What qualifies a system as a robot?

A system can qualify as a robot when it combines sensing, computation, and actuation, even if it does not resemble a walking mechanical person. Teller uses this definition for stratospheric balloons that sense conditions, rise or fall, and select winds. He also applies the concept to airborne wind turbines, self-driving cars, and self-flying package-delivery vehicles.

Q: How could biology shape future technology and engineering?

Biology could shape future technology through direct applications and as a model for complex engineering. The panel discusses cloud laboratories, de-extinction efforts, new life forms, and transplant organs designed not to trigger the immune system. Jurvetson also presents biology as inspiration for iterating on algorithms and compounding intelligence or complexity over time when creating future artificial intelligence systems.

Q: How should academic researchers choose future problems?

Academic researchers should begin with problems they care deeply enough about to commit their heart, body, and mind. Smolke argues that academia should look beyond a two-to-five-year horizon and consider needs 10, 15, or 20 years ahead. Researchers can then select foundational or applied work capable of producing transformative change, while drawing on the creativity of surrounding collaborators.

Summary

This is a panel discussion with Steve, Astro, and Christina about their perspectives on the past, present, and future. They discuss their past predictions, what they got wrong and what they got right. They also talk about the future technologies they find exciting, the importance of a broad education for engineers, and their responsibilities regarding societal implications of their work.

Questions & Answers

Q: What is one thing that didn't happen 10 years ago that you were sure was going to happen?

Steve reflects on several mistakes he made in predicting the future, such as thinking Amazon would be obsolete and that Netflix would fail.

Q: What did you predict 10 years ago that is happening now?

Astro shares his belief in the future of artificial intelligence and how it has continued to steadily progress in various areas.

Q: What is something that has happened and is changing the world that you couldn't have even imagined?

Christina mentions de-extinction as a topic that people are seriously working on, as well as the concept of cloud laboratories within biology.

Q: What is the coolest thing in the pipeline right now?

Steve discusses the transformations happening in various industries through deep learning and machine learning techniques. He also talks about the revolution of learning in life sciences and its impact on the future of engineering.

Q: How do you choose what to focus on and work on in the lab?

Christina emphasizes the importance of passion and choosing problems that will make transformative changes in the world. She also mentions the value of interdisciplinary teams and the freedom to explore new directions.

Q: How does policy play into future innovations?

Steve mentions the role of government in sponsoring projects but also states that they rarely rely on policy to make investments work. He sees government grants to universities as a way of helping the tech world.

Q: How do you educate the engineer of the 21st century?

Christina highlights the importance of teaching students how to choose what to work on and focusing on foundational research. She also mentions interdisciplinary collaboration and the need for broad literacy in biology.

Q: How important is domain expertise in today's rapidly changing world?

Astro believes that the ability to learn quickly and adapt is more important than specific domain expertise. He encourages the development of skills such as teamwork and communication.

Q: How do you think about linking technology, innovation, and societal responsibility?

Steve discusses the responsibility of engineers to be transparent and educate the public sector about the implications of their technologies. He also emphasizes the importance of anticipating and solving future problems.

Q: To what extent does technology drive our value system?

Steve discusses how technology can have an impact on our values and societal norms, but also mentions the importance of investors aligning their values with the technologies they support.

Takeaways

The panelists emphasize the need for continuous learning and adaptation in a rapidly changing world. They discuss the importance of interdisciplinary collaboration and the responsibility of engineers to consider the societal implications of their work. They also highlight the importance of transparency and education in shaping the future. Overall, they see technology as a driver of progress and innovation, but acknowledge the need for careful consideration of ethical and societal impacts.

Summary & Key Takeaways

• Past forecasting errors reveal how difficult it is to predict which companies, products, and industry transitions will succeed. Steve Jurvetson underestimated Amazon and Netflix, while Astro Teller dismissed camera phones. Teller argues that most ideas are wrong, so progress depends on searching through many possibilities to identify the few valuable ones.

• Machine learning is transforming fields that were previously dominated by specialized industrial hardware. Rockets, drones, robots, satellites, energy systems, and autonomous vehicles can increasingly combine off-the-shelf components with sensing, computation, actuation, and sophisticated control. This shift makes intelligence embedded in products a reusable approach across many sectors.

• Biology offers both practical opportunities and intellectual models for future engineering. The panel discusses cloud laboratories, de-extinction research, engineered life forms, transplant organs, and biological inspiration for building complex systems. Academic researchers should pursue problems they deeply care about while developing foundational technologies with horizons extending 10 to 20 years.