Speaker A
Let me start with some good news.
Sam Altman discusses the transformative impact of Codex and AI agents on knowledge work, software design, and future social interactions.
Key Takeaways
- Codex is a breakthrough AI model driving rapid innovation in software development and knowledge work.
- AI agents with broad system access will transform productivity and workflows but require new security and usability paradigms.
- Software design must evolve to accommodate AI collaboration alongside human users.
- Future social interactions may be mediated by AI agents, creating novel collaborative environments.
- Infrastructure and legal frameworks are critical constraints that need addressing for AI's full potential.
Summary
- OpenAI is the first design partner for Codex, which is rapidly being integrated into products like AI Defense.
- Codex represents a major advancement in AI, enabling economic value creation and transforming how companies and individuals work.
- The concept of full AI companies, where AI builds and operates complex software and interacts with the real world, is a key future vision.
- AI agents with full control of computers and web browsers enable powerful new workflows and productivity enhancements.
- Future social networks may involve multiple AI agents interacting on behalf of users, creating new forms of collaboration and idea generation.
- Obvious constraints include energy, hardware manufacturing, and compute capacity, while non-obvious constraints involve security, data access, and software redesign.
- Software needs to be rewritten to be usable by both humans and AI agents, potentially changing software architecture fundamentally.
- Current software often does not support simultaneous human and AI use, causing workflow disruptions.
- Legal, permissioning, and usability frameworks must evolve to support always-on AI computing, such as AI listening to meetings and managing data responsibly.
- The ongoing rapid advancement in AI will continue to impact science and various industries profoundly.
Full Transcript — Download SRT & Markdown
Speaker A
I don't know if you know this, but we are the first design partner for Codex.
Speaker B
I did know that.
Speaker A
Thank
Speaker B
you.
Speaker A
And we know in the past few months, I think we've hit an exponential curve.
Speaker A
AI Defense was a product that we launched over here last year.
Speaker A
In about two weeks or three weeks, 100% of the code in AI Defense will be written with Codex.
Speaker B
That's unbelievable.
Speaker B
I Codex has been my biggest update on AI in a while.
Speaker B
Uh, the the app that we launched yesterday, it just pushed things over the edge for me of where it's I'm like, all right, this is going to create an unbelievable amount of economic value extremely quickly.
Speaker B
This is going to change how all of OpenAI works, going to change how other companies work.
Speaker B
Um, it
Speaker B
the models really hit some threshold and I think now the interface and harness has caught up and
Speaker B
I think this is the first this you're talking about ChatGBT moments.
Speaker B
I think this is like the first time I felt another ChatGBT moment
Speaker A
for
Speaker B
here is a a clear, clear glimpse at the future of knowledge work and how enterprises and individual people are going to use AI to work in a completely different way.
Speaker A
So what's the upper limit on this, do you think?
Speaker B
Um,
Speaker B
I mean, the upper limit I think is like full AI companies.
Speaker B
There's probably an upper limit beyond that.
Speaker B
The current one I can think about is full AI companies.
Speaker B
And that seems very powerful.
Speaker B
Like this the idea that a coding model can create a full complex piece of software, but also interact with the rest of the real world to build a company around it is a very big deal.
Speaker A
And then this this notion of what's happening right now, I'm sure you're tracking this with Multibook and um, what happened with Cloudbot?
Speaker A
Like is it just a passing fad or do you think that there's something that we should take away from that on um,
Speaker B
No, I think it is definitely not a passing.
Speaker B
Well, Multibook may be.
Speaker B
I don't know, but uh, Openclaw is not.
Speaker B
Um,
Speaker B
I think this idea that code is really powerful but code plus generalized computer use is even much more powerful is here to stay.
Speaker B
The fact, I when I initially installed Codex, I said I was never going to give Codex full control of my computer with without checking what it was doing, and that lasted about like two hours because it was so useful.
Speaker B
And then I and then I kind of got persuaded by people that I really shouldn't have it running like that.
Speaker B
So now I have two laptops until I figure out how this is all going to work.
Speaker B
But giving an AI agent full use of your computer and your web browser with all your sessions leads to incredible stuff.
Speaker B
And that seems here to stay.
Speaker B
The the idea that we can take a Codex style workflow and expand it to all knowledge work.
Speaker B
Um, so that whatever you're doing, Codex can be or some version of that can be using your computer and using the web for you and editing documents and
Speaker B
whatever other kind of work you need, um, feels like a genuine transformation in how knowledge work will happen.
Speaker B
And that I think Openclaw did an incredible job of bringing many ideas together to make that feel usable and real, and that seems certain to be part of our future.
Speaker B
I think Multibook is cool, uh, and I think it points to something that will be real, and maybe it will be Multibook.
Speaker B
Uh, but there will be new kinds of
Speaker B
social interaction where you have many agents in a space interacting with each other on behalf of people, uh, leading to all sorts of new things.
Speaker B
Like you you can imagine a totally new kind of social network where everybody makes an agent or many agents and puts them in there and the agents are talking and doing stuff and finding them people and information and collaborating with other people's agents to come up with new ideas.
Speaker B
That I'm sure there will be an interesting.
Speaker B
I think the future of social may look something like that, very different than today.
Speaker B
Whether it's Multibook or not, uncertain.
Speaker A
Sure. And so if you think about the constraints that we are facing today.
Speaker A
You know, besides the obvious ones, the infrastructure and the compute and the power and all of those pieces.
Speaker A
Are there any non-obvious constraints that are actually going to that are holding us back because you know, people always say in the short term, you always overestimate the impact of these technologies and longer term it's going to be grossly underestimated, but what what are the non-obvious constraints that you see right now?
Speaker A
Like you're like, man, I wish if I had a magic wand, if I changed that.
Speaker B
So I think the obvious constraints are the biggest ones, uh, still, energy,
Speaker B
manufacturing enough hardware, all that kind of stuff.
Speaker B
The non-obvious ones that are most top of mind for me.
Speaker B
One, how how are we going to balance the sort of security and data access versus the utility of all of these models.
Speaker B
Um, I don't think anyone has a great answer to this yet.
Speaker B
I it feels to me like there is a new kind of security or data access paradigm that needs to be invented for this.
Speaker B
Um, another is how are we going to rewrite all software to be equally usable by humans and AIs?
Speaker B
There's like a bunch of weird quirks right now about
Speaker A
And does that change the architecture of the software itself where you're going to optimize it for agents more so than humans and so it fundamentally changes how you build software?
Speaker B
Yes, and there's like big examples of that and then there's like dumb examples of that.
Speaker B
So, for example, uh, I
Speaker B
I would love my agent to be able to use Slack on my behalf because I hate joining in Slack and I think it's like this chaotic mess for me.
Speaker B
And but it's important.
Speaker B
Uh, but the way it works right now is my agent can use the Slack web interface and go read all my threads and do something for me, but then it has marked a bunch of stuff as red in the process of doing that and it's broken my workflows.
Speaker A
Broken your workflows,
Speaker B
yeah. So I don't like that's just a silly example of how software is not like a lot of software is not quite meant for an AI and a person to be using it together.
Speaker B
Um, maybe we'll want AIs to like have different
Speaker B
sort of user accounts in some ways using the same kinds of things.
Speaker B
Maybe a lot of software will get rewritten so that it's primarily or largely used by AI, but also still works for people using it the old-fashioned way.
Speaker B
Um, another kind of non-obvious block is how like one of the most powerful things about AI is you can do this sort of always-on computing where you could have an AI um, listening to your meeting or watching your meeting and you know, watching what you're doing on your computer and uh, then just like add a lot of value and do stuff for you.
Speaker B
We don't even our like existing computer hardware is not really meant for that.
Speaker B
Our permissioning system and how we think about what an AI gets to see and do stuff with and what it gets to keep is not really meant for that.
Speaker B
Um, our legal system doesn't really support that.
Speaker B
Well, you'd like to be able to, you know, record a meeting and learn something from it and delete the recording.
Speaker B
Uh, so I think there's a lot of just usability things like that.
Speaker A
And then, you know, you talk about like one of the things that I find is a huge dichotomy is there's so much advancement that's happening in science and in um, in all the different
Topics:CodexOpenAISam AltmanAI agentsknowledge worksoftware designAI securityAI collaborationfuture of workAI social networks
