“There is a book I recommend everyone interested in the human-AI future read. Max Bennett’s, ‘A Brief History of Intelligence,’ 2023, supports a claim I’ve long held – the only way through to advanced, trustable, secure, agentic AI will be by recapitulating the intelligence (both intuitive and deliberative), emotion (which solves incessant logical impasses in human thinking), immunity and the deeply prosocial yet also deeply competitive ethics and instinctual algorithms previously discovered and programmed into us by evolutionary development.
“Bennett’s book makes clear how incremental the AI improvements will be over the next 10 years, even as the hype and funding grow to gargantuan levels.
AI in these still-early years will remain mostly top down, benefitting powerful actors and holders of capital. … Sadly, the economics of making personal AIs don’t work in a world in which AIs are still not agentic and where there is deep mistrust in them and pessimism for our societal future – a consequence of plutocracy and accelerating change. … Neuroscience and genetics still have many secrets to be uncovered before we’ll have truly self-improving AI, and that AI – when it arrives – will be a new form of life, with its own agency, yet one that is also deeply connected to and ethically aligned with us, at least with our sentience and complexity protecting and promoting values and virtues.
“Neuroscience and genetics still have many secrets to be uncovered before we’ll have truly self-improving AI, and that AI – when it arrives – will be a new form of life, with its own agency, yet one that is also deeply connected to and ethically aligned with us, at least with our sentience and complexity protecting and promoting values and virtues. Meanwhile, we stumble along.
“AI in these still-early years will remain mostly top down, benefitting powerful actors and holders of capital. But, as it grows, decentralized and personal forms will also emerge. I’ve long written about the advent of personal AIs (PAIs), with private data models, easily modified via conversation with our AI agent. Sadly, the economics of making personal AIs don’t work in a world in which AIs are still not agentic and where there is deep mistrust in them and pessimism for our societal future – a consequence of plutocracy and accelerating change.
“Khanmigo, a beautiful example of how to use AI to enhance individual thinking skills, is presently facing strong adoption headwinds due to both institutional and public fear, uncertainty and doubt over the use of this new technology. Inflection’s Pi, an AI helping with empathy and kindness, lost its leadership to Microsoft to pursue more lucrative AI aims. AI will have to get a lot more powerful to overcome these adoption and economic barriers. The beautiful visions of the future described in Sal Khan’s ‘Brave New Words’ 2024 (the best new book on the future of AI for education and job training) will arrive only for a privileged or courageous few over the next decade.
I fear that while there will be a growing minority benefitting ever more significantly with these tools most people will continue to give up agency, creativity, decision-making and other vital skills to these still-primitive AIs and the tools will remain too centralized and locked down with interfaces that are simply out of our personal control as citizens. … We’re still walking into an adaptive valley in which things continue to get worse before they get better. We will experience too much ‘Wall-E’ and not enough ‘Incredibles’ in our next 10 years, to be sure.
“I fear, for the time being, that while there will be a growing minority benefitting ever more significantly with these tools most people will continue to give up agency, creativity, decision-making and other vital skills to these still-primitive AIs and the tools will remain too centralized and locked down with interfaces that are simply out of our personal control as citizens.
“It will finally have arrived when you can permanently ban an ad for a drug, gambling, car or any other product or service from your personal view screens just by talking to your Personal AI (PAI).
“When you can complain about any product or service – at point of use – and have that go to the public web (or a private database if you accept the discount) when your PAI is advising you on boycotting, initiative politics and UBI reforms. Then it will have finally arrived as I would define it. All else will be just more distracting circuses, not sustaining bread.
“I fear we’re still walking into an adaptive valley in which things continue to get worse before they get better. We will experience too much ‘Wall-E’ and not enough ‘Incredibles’ in our next 10 years, to be sure.
I would bet the vast majority of us will consider ourselves joined at the hip to our digital twins once they become useful. … [if] we have the courage, vision and discipline to get through this AI valley as quickly and humanely as we can.
“Looking ahead past the next decade, I can imagine a world in which many of us are running lifelogs that capture and use our conversations and experiences; a world with trusted PAIs with private data models (as private as our email, text and photos) that the marketers and state don’t have direct access to (except under subpoena); a world in which our PAI knows us well, looks out for our values and goals, educates our kids in the way Sal Khan hopes, and continually advises us on what to read, watch and buy, who to connect with to accomplish our goals, what goals are most useful to our passions, abilities and our economic status.
In a world in which open-source PAIs are among the most trustworthy and human-centered many political reforms will re-empower our middle class and greatly improve rights and autonomy for all humans, whether or not they are going through life with PAIs. I would bet the vast majority of us will consider ourselves joined at the hip to our digital twins, once they become useful enough.
“In the meantime, and on average for the next decade at least, I expect PAIs will be only weakly powerful and weakly adopted and the divide between ‘lean forward’ AI users (growing their knowledge, productivity and soft skills) and ‘lean back’ users (sliding further backward on many of our most precious human traits) will only grow. I hope we have the courage, vision and discipline to get through this AI valley as quickly and humanely as we can.”
This essay was written in January 2025 in reply to the question: Over the next decade, what is likely to be the impact of AI advances on the experience of being human? How might the expanding interactions between humans and AI affect what many people view today as ‘core human traits and behaviors’? This and nearly 200 additional essay responses are included in the 2025 report “Being Human in 2035.”