AI and surprise do not go well together: Sam Altman on how startups should build long-term value
And are open-source model a danger to OpenAI biz?
In a detailed discussion with RedPoint venture’s Logan Bartlett, Sam Altman, CEO of OpenAI, shares his insights on the future of artificial intelligence, the launch of ChatGPT-4o, and the implications of AI advancements on society.
This is his first interview after the launch of GPT-4o and he delves into the significance of generalized AI models, discusses the importance of AI regulation, and envisions a future where AI-human interaction is facilitated through humanoid robots.
Here are the big ideas from the conversation
"Human adaptability to almost anything is much more remarkably strong than we realize... you can get used to anything as the new normal, good or bad, pretty fast."
The personal impact of running OpenAI
Altman reflects on the personal impact of running OpenAI, highlighting the loss of anonymity and the significant shift in his life over the years.
The launch of ChatGPT-4o
The launch of ChatGPT-4o, a multimodal AI model, is crucial for its naturalness in voice control and seamless integration of text, voice, and vision capabilities.
Are the open source models a threat to OpenAI business?
"No current one does, but I could imagine one could” - Sam Altman
What about the safety issues that most open source models brag about? Safety is not a binary thing.
Advice for founders amidst AI advancements
You can either build a business that bets against the next model being really good or a model that bets on happening and benefits from it happening.
If you are doing a lot of work make one use-case really work that's just beyond the capability of GPT-4 / GPT4-0 and you get it to work, then GPT-5 comes up and it does that and everything else really well - you are sad about the effort you put into that one thing to get it barely work but if you had something that just like kind of worked across the board and people were finding things to use for but you didn't put in like tons of effort to make this one thing kind of possible and GPT-5 comes along and that makes everything just way better - then you have got rising tides lift all your boats effect.
In most cases, you are not building an AI business, but you are building a business and AI is a technology that you use.
The need for generalized AI models
Altman emphasizes the need for generalized AI models capable of true reasoning over deeply specialized ones.
Humanoid robots and AI-human interaction
He envisions a future where AI-human interaction occurs through humanoid robots, facilitating effective communication through natural language.
CEOs are advised to focus on the long-term defensibility of their businesses amidst AI advancements.
"We should all hope for a world where intelligence is too cheap to meter... and you don't even think about whether... it's wildly abundant."
With intelligence as a service becoming increasingly accessible annually, the emergence of new job titles centered on human connection and unique experiences signals a shift in the job market due to AI progress.
Importance of interpretability in AI models
Altman underlines the importance of interpretability in AI models, advocating for a deep understanding of neural networks and logical flaws in their outputs.
He addresses the balance between excitement for AGI development and concerns about control and oversight of powerful AI technology.
"AI and surprise do not go well together."
The concept of AGI as a continuous progression
The concept of AGI is viewed as a continuous progression rather than a sudden discontinuity, requiring ongoing vigilance and progress monitoring.
Research in AI progresses unpredictably, with breakthroughs sometimes occurring faster than expected, showcasing the dynamic nature of AI advancement.
Importance of AI regulation and oversight
Altman discusses the importance of AI regulation and oversight, considering concerns about potential risks associated with AI development.
He delves into safety in AI, drawing parallels to safety standards in industries like aviation and medicine, emphasizing the need for a preparedness framework.
"I think many years from now, humans are still going to care about other humans... we're so wired to care long term about other humans in all sorts of big and small ways."
OpenAI's approach and philosophy
OpenAI's approach to decision-making, particularly regarding the iterative deployment of AI technologies, stands out as a departure from conventional methods.
This strategy of openness and collaboration deviates from the secrecy often associated with AGI development, showcasing a unique philosophy.
Altman stresses the value of promoting internal talent while also bringing in new senior executives at OpenAI to avoid reinforcing a monoculture.
He envisions a future where AI assistants are viewed as separate entities from individuals, emphasizing the importance of maintaining distinct interactions.
Altman advocates for decentralization in AI development, promoting collaboration without centralized control, and predicts significant advancements in productivity and innovation through AI integration in various fields.