The Real Implications of Generative AI
What we learned in an amazing Meeting of the Minds at Ground Zero in San Francisco
The explosive arrival of Generative Artificial Intelligence in the last six months has generated a vast number of questions, starting with big picture ones: What is really going on with this new evolution of AI? How big a deal really is it? What’s probably coming in the near-term, like in the next few years or through the 2020s? What’s now possible to achieve with this general purpose AI over the long-term, as it scales up over the next 25 years? And then what should we be doing now to make sure we bring out the best of AI and avoid the worst?
Then if you step down from those big picture questions, you confront another level of questions: What are the real implications of Generative AI? What are its real technological capabilities right now and in the next few years? What are the implications for business, and which industries will be impacted first and which remain relatively immune? How could this boost general productivity rates or even growth rates in our economy over time?
Then what are the positive possibilities of Generative AI? What is the potential for this new technology to go beyond business and transform other fields in our society for the better like education, or healthcare, or democracy itself? Could we use this powerful new tool to solve many of our intractable challenges like climate change? But then what are the actual risks of Generative AI? Not the hyped up risks that are proliferating through the media, but what are the actual risks that we need to address to responsibly move forward?
So there’s plenty of important questions, but what about the answers? No one person knows the answers to all of these questions, or arguably any one of them at this early stage. At this point the best you can do is try to piece together answers by bringing together many different people who may hold some piece of the puzzle.
“No one is as smart as everyone.” I attribute that quote to Kevin Kelly, the founding executive editor of WIRED magazine who hired me to work with him at the beginning of the last tech revolution of this scale, the digital revolution that opened up the internet to everyone in the early 1990s. WIRED arguably acted as that convener of many of the people who held a piece of the puzzle and together could help explain how these strange new startups and the nascent digital economy could evolve over the next 25 years, and mostly did.
So when Generative AI essentially opened up the world of AI to everyone this year, San Francisco felt like it was in a time warp. The city was energized by a flood of young technologists and entrepreneurs coming to what was nicknamed Cerebral Valley. And many of the veterans of past technology iterations were scrambling to figure out the full significance of this moment and what lies ahead. Everyone wanted to compare notes with their peers and check in with others from diverse networks to piece together what was happening.
So my company Reinvent Futures launched a new event series called The Great Progression that kicked off with an initial Meeting of the Minds to first help answer the question: What are the real implications of Generative AI? We partnered with the hip new club Shack15 housed in the Ferry Building, which literally is ground zero in San Francisco, and so ground zero of this Gen AI revolution too. And then we partnered with Cerebral Valley, which has emerged as one of the main organizations connecting up the nascent world of new Gen AI startups and hackers.
We then sent out invites to some of the smartest, most knowledgeable, most innovative people I’ve come to know in more than 25 years working in Silicon Valley and the region. And we maxed out the space with 250 people, with another 250 on the waiting list. We had wise OGs come out like Tim O’Reilly, Mitch Kapor, Kim Polese, and Larry Brilliant. (Google them for more context.) We had founders of well-known companies like: Jane Metcalfe of WIRED and Linda Avey of 23andMe. We had leading academics from UC Berkeley and Stanford like Ken Goldberg and Brad DeLong. We had major media from places like The New Yorker and Bloomberg too.
We had top companies want to get involved in sponsoring and bringing in experts from their networks who wanted to learn too: We had Mike Haley, head of Research from the software company Autodesk. Marc Oost, the Global Leader of AI for the Paris-based consulting firm Capgemini, flew in from Amsterdam. And Brad Jackson, CEO from the consulting firm Slalom came down from their headquarters in Seattle.
Clearly people wanted to talk out these questions and compare notes with their peers in order to start to collectively formulate answers. But the people who jammed into this elite gathering in the heart of San Francisco are not the only ones looking for answers. So this Substack series of my written essays are a way to share what we are learning more broadly.
I will synthesize some top-line observations from each event we do, and pass on some of the best insights from some of those we have asked to briefly speak in our program. Unfortunately, much of the learning also comes from the many private conversations during the party after the program, which goes on for hours after the event. For that, you’ll have to come join us.
I’ll start my top-line takeaways with a caveat. No one person can perfectly synthesize the views of a dozen speakers with the full nuance of their individual insights. And that’s even more so for the rest of the 250 people in the room. Plus I have my own biases that tend to make me more positive about the developments of new technologies, more can-do about how we can deal with any issues, and generally I’m an optimist by nature.
Generative AI is a Very Big Deal
One of the first things everyone wants to know is how big a deal is Generative AI? Pretty much everyone in our Meeting of the Minds confirmed that this was a very big deal. Many people thought the arrival of Generative AI was at least as big a deal as the arrival of the Internet in the early 1990s, and probably bigger.
That analogy is worth elaborating on because in each of those two cases the significant time was when the technology became available to everyone. The internet had begun in the 1970s and was used by segments of the military, government and academia for years before the 1990s. But for all practical purposes people consider the internet to have arrived in the 1990s because the World Wide Web made it accessible to everyone.
Generative AI did the same thing for Artificial Intelligence. Earlier forms of AI had been around for decades but it was only accessible to governments and corporations with significant resources and technical expertise. The arrival of Gen AI with ChatGPT in November 2022 opened it to everyone. And within just two months 100 million people around the world had become monthly users.
The person who really drove home the big deal point was Ken Goldberg, Chair of AI and Robotics at UC Berkeley and Cofounder of Ambidextrous Robotics, a startup that develops software for AI-based robotic picking in e-commerce applications. Despite his deep expertise, Goldberg has long been skeptical of how much AI and robots could ultimately do in his lifetime. Goldberg has been in the AI field for 40 years and has watched all the ups and downs of broken promises, and did not think in his lifetime AI could pass the Turing test, which means could pass as a human in a conversation with a real human.
The arrival of Gen AI has fundamentally changed Goldberg’s mind. He now thinks with a few tweaks AI will soon pass as human. What’s more, he always thought that AI could never be creative like humans. Now he sees how Gen AI already is creative and will get more creative with time. (You can see extended quotes that give more context from Goldberg and others referenced here in another Substack article on Extended Quotes coming soon.)
Goldberg talked about how much this breakthrough has shaken his world view. He made the analogy to what happened 600 years ago when Galileo first posited that Earth was not the center of the universe, challenging the orthodoxy of that time, but ultimately forcing a fundamental rethink of what it meant to be human.
“I think we're at a similar moment where there is a real reckoning of maybe there's another form of intelligence,” Goldberg said. “I'm not afraid that it's going to take over and dominate us, but it's a different form of intelligence and I think it's incredibly interesting and it's an opportunity for us to expand our own minds, to create new things, and we are going to be able to figure this out.”
A Huge Threshold in the Evolution of Computing
One of the great OGs of the tech world from the last 40 years explained the historical significance of the Gen AI breakthrough by framing it as crossing a huge threshold in the evolution of how computers interact with humans. Tim O’Reilly is the founder of O’Reilly Media that caters to software coders, the curator of the exclusive FOO Camp unconferences, and the person who coined the concept Web 2.0 through his conferences.
O’Reilly talked about how each great advance of computing closed the gap between how machines work and how humans work, in other words how the interface between the two got us closer to human speech. He then walked us through the evolution of computing from arcane early computer languages through progressively easier coding languages, and from mainframe to personal computers to the web and to more intuitive mobile interfaces.
“And guess what? This Generative AI is the next big wave,” O’Reilly said. “For the first time, we really are starting to have computers that have gotten smart enough that they're coming all the way towards us where we can actually speak with them in our language and they can understand it. So that's a profound shift. I think we're just at the beginning of an astonishing new wave that really is bigger than any of the previous waves, just as each one that came along was bigger than what came before.”
O’Reilly was most excited by the potential for Gen AI to give us a super tool that can help us solve many of the most daunting challenges of our times by giving us the ability to deal with complexity through coordination that is beyond human abilities alone.
“There's a great quote from a guy named Paul Cohen, a professor at University of Pittsburgh, who once said, ‘The opportunity of AI is to help humans model and manage complex interacting systems.’ If you look at the challenges of the world today — climate change, economics — these are challenges of coordination.”
Gen AI is Opening up Corporate AI for All
We heard from quite a few people with expertise from inside the corporate world who had been getting early access to earlier forms of AI. They could help others understand what now might be available to all, as well as what the more general models could help corporations do now too.
“Even before the announcement of ChatGPT back in November of last year, I was a believer that artificial intelligence is the most transformative technology of our generation,” said Michelle Lee, former head of Amazon Web Services Machine Learning Lab, as well as the former Director of the US Patent Office.
Lee said she and her teams within Amazon had been working with companies and government agencies like NASA to apply AI to relatively narrow proprietary databases to predict outcomes that humans alone could not foresee, like the probability of congestive heart failure a year before it would occur, or advance warnings of solar superstorms.
But Generative AI has ingested roughly everything on the internet, 45 terabytes of data, and so has a generalist capability that no one has had until now. “So this generative AI technology will absolutely increase all of our productivity,” Lee said. “It will be an accelerant and an augmenter of all of our capabilities.”
The Practical Side of What Business Can Now Do
What was most useful in this Meeting of the Minds was how each speaker with just 5 minutes could build on the insights of the previous speakers. Arjun Prakash, CEO and Cofounder of Distyl AI, a new Generative AI company that helps companies take advantage of OpenAI, built on the previous insights to explain in more detail just what has changed with Gen AI.
Most companies in the world have spent the last decade or more going through what the business world calls the digital transformation, which is converting all data to digital form. But up until now any organization had to rely on data scientists or machine learning engineers to make sense of any of that data. But now anyone who speaks English can theoretically interact with that data without intermediaries.
“So all of a sudden we have increased the number of people at an enterprise that can communicate with an AI model,” said Prakash, who worked for a decade with Palantir, a machine learning company that often works with big governments. “Many people can now ask questions and get answers.”
Gen AI also dramatically speeds up the process of creating value from all that data. Previously, engineers had to build an AI model from scratch to extract what they wanted to learn from the data - and that took time. “Now I can start with something like GPT 4, which out of the box gives me the ability to get a working prototype in days, not months,” he said. “That's an incredibly fast time to value.”
And having an all-purpose model like GTP 4 allows you to use the same core model to solve many tasks that previously needed their own models, or engines so to speak. This can dramatically drop the costs. “I can now maintain just one model, a GPT 4 model for example, and instead have 200 different prompt templates for 200 different tasks,” he said. “I think this is going to be a game changer for enterprises.”
Independent Agents Soon on the Way
We’re just in the very early stages of this Gen AI revolution, similar to what took place in the initial Web 1.0 era with the arrival of the internet in the 1990s. Humans are the ones directly prompting the new AI for immediate answers. But within the next few years we will be entering the next stage that could be analogous to Web 2.0 where the AI will be able to autonomously act in the world.
“I think having a highly capable Executive Assistant is going to start to be possible in the next year or maybe two,” said Kanjun Qiu, CEO and Cofounder of Generally Intelligent, an AI research lab building general purpose agents that can be safely deployed in the real world. “This is not very far off.”
Qiu said this AI agent on the horizon would be able to act like a human executive assistant does now. It would be able to organize activities that take place outside the internet on its own, make simple decisions that align with your goals, and understand what was impossible to do and suggest revising those goals.
Within the 2020s, she expects that you will be able to speak English to the computer and explain what you wanted the computer to do, and then the computer might reconfigure its own software, create its own computer code, to be able to carry out your task. That’s a very different conception of computers.
Caution Warranted But Not Too Much
The general thrust of the conversation of the evening was overwhelmingly positive about the implications of Gen AI, but there were various notes of caution. Sunil Paul, a serial entrepreneur who has been CEO and Cofounder of several companies, including Brightmail, Sidecar, and Spring Free EV, warned all of us to be careful about making common errors in human judgment:
Be careful not to make The Acceleration Error, assuming that the pace of change and improvements will just keep going on without end. They might not, like what happened with nuclear power, or space exploration. All new technologies are embedded in human contexts that can develop constraints.
Or beware making The Breakthrough Error, which assumes that one breakthrough like this will be followed by others in the same vein. We fell into that trap when IBM made a breakthrough in manipulating individual atoms, but the field of nanotechnology never took off.
“Then there is the human analogy error,” said Paul. “In other words, because these things seem like us, they chat like us, they interact with us, we can make the mistake that they are like us, that they evolve, that they have agency.”
“They don't have agency, they do not evolve," he said. “They are agents of us. They are part of something that we are creating.”
The Concerns about AI are Really Concerns about Humans
Like many in the gathering, Tim O”Reilly did not express much concern for the dangers of AI that seem to be getting a lot of attention in the media. In fact, he stressed that some of the issues that people are most worried about are not problems with the technology but problems with human society.
GenAI should be considered more like a hybrid of machine and human, not an independent entity. GenAI is a large language model that is built upon massive amounts of written material from human beings.
“They're a mirror,” O’Reilly said. “They're a mirror of all that's good and bad in our society, and it's really important to remember when we think about, ‘Oh, we're going to fix the bias.’ We don't want to be fixing the mirror. We want to be fixing what it is showing us, which is us.”
I want to give a thanks to those whose support makes this ambitious project possible, including this essay. Our partners Shack15 club in the Ferry Building in San Francisco and Cerebral Valley, the community of founders and builders in Gen AI. And our sponsors Autodesk, Capgemini, and Slalom who help with the resources and the networks that bring it all together. We could not do this without you.
Fantastic! Thank you so much!