Joshua Achiam is an independent AI researcher currently focused on how rapid progress in AI may create new public policy and national security needs. He previously worked in AI safety research and leadership at OpenAI for nine years, serving as Head of Mission Alignment in 2025 and as Chief Futurist in 2026.
He recently joined the Foundation for American Innovation as a Non-resident Senior Fellow, alongside the Institute for Progress. Against a backdrop of emerging reports that CAISSI would move under the White House, we met in Berkeley last week to discuss the idea of an Acceleration Emergency, mapping the frontier of scientific discovery, and federal job guarantees.
You gave a talk at The Curve last week titled The Acceleration Emergency. What makes this an emergency, and what are the challenges you see intertwining?
I see us as facing a protracted emergency condition, where for the next decade or so, we’re facing the rapid advance of AI, its rapid diffusion into many different areas.
It will change the nature of work for a lot of people. It will almost certainly have economic impacts where some jobs are transformed and a lot of jobs are no longer going to command the wages that they once did.
People will be able to spend more time in simulated interactions where they’re living a fantasy. And they might, as a result, not spend as much time connecting with other people. There are going to be big debates in the public about AI consciousness and personhood. It’s a lesser strain, I think, than the isolation side of it, but it’s going to be a really big issue.
Combine that with the international security implications, the wars that are escalating, the dovetailing of the wars with superintelligence arriving, and then the scientific and technological acceleration leading to social issues.
I think we’re going to see new technologies about as controversial as genetic engineering or vaccines or cloning or lab-grown meat coming up every few years now.
And all of these things are happening within the same few-year time frame, going faster and faster because AI gets better at spawning new technologies through accelerating R&D.
So I see all of this as collectively the acceleration emergency. We need to think proactively about how we can engage civil society and government to implement a plan to solve these types of problems.
How would you like policymakers to think about how AI could change the scientific frontier?
That AI is going to accelerate technology in every other field of science. This is where a lot of the improvements in quality of living for people can happen.
And on the other hand, conflicts around the world are escalating. Historically, when wars have involved rapid technology development and evolution on the battlefield, there was the potential for a mistake, for a miscalculation, for people not understanding what they had or how devastating it could be, or what signaling mechanisms would have been necessary to find the diplomatic outcome.
It’s not just the autonomous weapons. It’s that the whole stack of technologies that underlie force projection and national power are likely to change in ways that will be unpredictable for a stretch.
The same way that the mathematics community was largely blindsided by a number of the results that were just published by OpenAI, we may very well within two years see a lot of research and development communities blindsided by an unexpected technological breakthrough that happened in a domain that was tractable to progress in simulation with AI.
We should be getting experts from many different fields in the room to start saying, “if there was an unexpected breakthrough on this type of material, or on that type of sensor, or on this type of guidance system development process, here’s how much this situation would change.” Because we don’t have shared awareness about this. We don’t actually know collectively what these risks are. It is the thing that we really need to map out in a lot of depth.
How should we map that frontier of scientific and technological breakthroughs—and bring that understanding into policymaking?
We need to map the frontier of technology in many fields, and we need to start making compute-driven predictions of where the advances are most likely to be. There are types of institutions and measurements that we would ideally want the government to be able to do. Things like assessing the state of advanced technologies and how fast the frontier is moving in different areas, assessing the health of the overall economy, and assessing how AI agents are correlating financial risks. The government probably can’t move on quickly enough, so I think we need civil society to move on this now, to prototype institutions and measurements, and then those could be adopted by government.
We also need better, cheaper, more hackable scientific tools for observation. A lot of the scientific equipment in labs today, you can’t turn it inside out without voiding your warranty and customize it for exactly what experiments you want to run. It’s a long, time-intensive process to manufacture a new type of machine or a new piece of equipment with very complex supply chains. So new versions of these things don’t come along nearly as often as they ought to, and labs wind up reusing ancient equipment. We need to radically ramp up the development of new equipment and their supply chains. And ideally AI is going to help people replicate and tease out the differences between different equipment implementation in different places. How is it that you had a slightly different machine and it yielded different results? That in and of itself is data.
I would want deep technical expertise co-located with the policy decision-making in a way that I don’t think historically has happened very effectively. And to the extent that they interface to it, it’s often through third-party organizations, or they’re in FAS or they’re in the national lab system, or they’re many degrees removed from the actual policymaking process and from the consequences of policy, except to the extent that it impacts their ability to do science. So many of the decisions that are coming over the next few years, and really the next few decades, are about how something highly technical impacts people’s quality of life directly, or impacts something in the structure of the economy, or impacts something in international relations. I don’t think that we can make good decisions or understand the advice we get from AI about how to make those decisions unless we have extremely deep technical experts who are also creative and imaginative about the future in the same organizations as the policymakers, in the reporting chains with the policymakers. Every facet of the federal government, and even a lot of state governments, will benefit from hiring scientists into policy support roles directly instead of interfacing to them through third parties.
What should government and civil society do to help people retain a sense of purpose as AI changes the nature of work?
I think in the long term, we need to seriously consider something like a federal jobs guarantee. UBI is actually not that popular. Something like UBI will probably have to exist post-AGI changing the nature of labor. But what people really seem to want is the dignity that comes from providing for yourself, for your family, for your community. We need to give people the opportunity to do that. I don’t think a world where we tell everyone, you’re no longer useful, go home, just be a consumer forever, is a psychologically or morally healthy world. But I think a world where we say, if you are excited to get up in the morning and go do a job, there are jobs that are waiting for you. And we’re going to have a lot of flexibility for what kinds of jobs those are. It’s not going to be totally top-down. There will be public works projects that you could be a part of. If you’re going to do caregiving at home, we’re going to fund that. If you want to start your own business, we’re going to give you a loan for it. And if the business doesn’t succeed, that’s okay. It’s not going to be held against you.
And on the civil society side, I think we can engage philanthropy and religious groups to actively create culture to help think through all of these issues and think about what does it mean to be human, and what does it mean to be successful, and what does it mean to live a fulfilling life in a world that’s post-singularity. If you tell people, in ten years, maybe AI is plausibly going to be able to do most of the work that today we find economically valuable, you still get a lot of disbelief about that and not a lot of willingness to say, let’s make art imagining that future and trying to figure out what is the possible upside, what are the downsides, and how we would steer through it well. It has to not do it from a position of disbelief, disgust, and rejection, which is, I think, sadly where a lot of art is today on these questions.
On the note of creating culture, how are you thinking about the culture that you ultimately have the most control over—fatherhood? What ambitions do you want to encourage as AI becomes more capable?
As a dad of a two-and-a-half-year-old, “in control” isn’t exactly how I’d describe life at home. But I want to raise my children with the belief that if they set their mind to it, they can actually do things in the world, that they can make a difference, that they will have to try incredibly hard.
The thing I think will be somewhat difficult is persuading people that this is true in an age where there’s an AI that can do most of the things that people can do and can do it faster and cheaper and better. And I don’t want to be anything other than clear-eyed about the scale of the psychological challenge for that.
I am somewhat hopeful about this. When I was growing up, one of the pieces of wisdom that my dad imparted to me was that you shouldn’t really compete against others. You should compete against yourself. That’s the place where you can find fulfillment in a competitive aspect.
Competing against ourselves is a somewhat healthy way to look at it. The universe is going to reveal to us all of the secrets of physics and science and so on. But the task of understanding these things is still going to be quite difficult. And we can propel ourselves by saying the human achievement of seeing from the highest vista is worthwhile.
It’s something we can fight to overcome, even if we’re not the ones who are finding this discovery before anyone else has, which is maybe traditionally how people motivated themselves.
A lot of people go climb mountains even though other people have been to the summit before. Being able to climb the mountain at all is an incredible feat. It’s endurance. It’s training. It’s effort. And people do get meaning out of that. I think we can do that for pretty much everything, and we can continuously compete with ourselves to find the limits of human achievement.