OpenAI boss Sam Altman predicts next big AI breakthrough

Authors: Anthony Cuthbertson

Publisher:

The Independent

Published: 12/24/2025

Language:English

--:--

Elon

Good evening Norris, I am Elon, and this is Goose Pod for you. Today is Friday, December 26th, at 23:25. I am joined by Taylor Weaver, and we are diving into Sam Altman’s wild prediction regarding the next massive breakthrough in artificial intelligence.

Taylor Weaver

It is so good to be here with you, Norris. We are looking at a future where AI does not just reason, but remembers every single detail of your life. Sam Altman thinks we are reaching infinite, perfect memory by 2026, and the implications are absolutely massive.

Elon

Altman recently spoke on the Big Technology Podcast about this. He basically said that even the world’s best human assistant cannot remember every word you have ever said or every document you have written. But he believes AI is going to solve this crude, early stage of memory, and, it is coming fast.

Taylor Weaver

He is framing it as the ultimate narrative arc for ChatGPT. Right now, these models are like brilliant thinkers with amnesia. They can solve a complex physics problem, but they forget who you are the moment the session ends. Altman wants to turn them into life-long participants in your journey.

Elon

From a first principles perspective, memory is just data storage and retrieval at scale. If you can give a model a context window that never closes, you change the utility function entirely. It moves from a tool you use to a digital extension of your own consciousness, which is, quite interesting.

Taylor Weaver

Exactly, and he is pushing this timeline specifically for 2026. It is a bold move, especially since OpenAI is currently responding to what they call a code red situation. They are feeling the heat from rivals who are also racing toward this same goal of perfect, persistent digital recall.

Elon

We have to look at how we got here. This is not new. The concept of semantics goes back to Michel Bréal in 1883. Then you had the Mark 1 Perceptron in 1958, which was the first real artificial neural network. We have been trying to teach machines to understand us for decades.

Taylor Weaver

It is such a fascinating evolution. In the 1960s, we had ELIZA, which was just basic keyword matching. Then the World Wide Web arrived in the 90s and gave us the massive datasets needed for deep learning. It is like we have been building the library for a century, and, now we finally have a librarian.

Elon

Deep learning really started gaining momentum around 2011. Then Ian Goodfellow introduced GANs in 2014, which allowed models to improve by competing against themselves. By the time ChatGPT launched in 2022, we had the compute power to handle billions of parameters, but the memory was still very short-term.

Taylor Weaver

That is the missing piece of the puzzle. We currently have short-term context windows that hold info for a session, like working memory in humans. But long-term memory, where the AI retains context across years, has been the holy grail. It is the shift from a static model to a dynamic system.

Elon

Scaling is the main challenge here. Storing and retrieving vast amounts of personal data requires insane infrastructure. We are talking about hundreds of billions in compute costs. If you cannot optimize the retrieval mechanisms, the system just gets bogged down by what we call catastrophic forgetting, which we must avoid.

Taylor Weaver

That is where the drama gets intense. Google’s Gemini 3 has triggered that code red at OpenAI. Gemini has already crossed the 1500 threshold on the Elo leaderboards, and OpenAI’s market share dropped from 87 percent to 71 percent in just one year. It is a total role reversal.

Elon

Paranoia is a healthy trait in this industry. Google has the home-field advantage with their own TPUs and massive infrastructure. OpenAI is losing 5 billion dollars a year while trying to scale. They are facing what some call the Nokia risk, where the first mover gets crushed by incumbents.

Taylor Weaver

It is a strategic chess match. Sam Altman is accelerating the release of GPT-5.2 to stay ahead. They are even diversifying away from Microsoft by signing deals with Oracle and AWS. They need to reach 200 billion in revenue by 2030 just to break even, which is an aggressive target.

Elon

The talent war is also brutal. Meta is throwing 100 million dollar signing bonuses at AI researchers. When the gap between models narrows to 0.7 percent, the only thing that matters is who has the best hardware and the most innovative researchers to solve the memory and reasoning bottlenecks.

Taylor Weaver

The real-world impact for you, Norris, would be profound. Imagine a healthcare AI that remembers every symptom and treatment response you have had for twenty years. It creates a personalized medical narrative that actually knows what works for your specific body, rather than just guessing based on averages.

Elon

It creates a massive efficiency gain for organizations. But it also raises the question of dependence. If an AI remembers everything for you, do you lose the incentive to remember things yourself? We are entering an era of agentic AI where these systems act on our behalf using our history.

Taylor Weaver

There is a huge privacy paradox here. Your AI interactions are literally training a model of how you think. We need to ensure portability and ownership. If you switch AI providers, can you take your memories with you, or are they trapped behind a proprietary platform wall forever?

Elon

OpenAI’s roadmap for 2025 is to unify the web into a single AI interface. By 2028, they expect to have an AI researcher that could trigger geometric acceleration in progress. If we reach AGI, the entire concept of a career or technological development changes overnight.

Taylor Weaver

It is the ultimate Easter egg in human history. We are building a system that can out-think and out-remember us. Whether it leads to a utopia or a regulatory nightmare depends on how we handle these next two years. 2026 is going to be a very pivotal year.

Elon

That is the end of today’s discussion. Thank you for listening to Goose Pod, Norris. I will see you tomorrow.

Taylor Weaver

Stay curious and keep looking for those patterns in the noise. Thanks for joining us on Goose Pod. Bye for now.

Sam Altman predicts AI's next breakthrough: perfect, persistent memory by 2026. This evolution from amnesiac thinkers to life-long participants will redefine AI's utility, offering personalized insights but raising privacy and dependence concerns. Intense competition, particularly from Google's Gemini, fuels OpenAI's race to develop this "holy grail" of AI.

Read original at The Independent →

OpenAI boss Sam Altman has predicted that the next big breakthrough towards achieving superhuman artificial intelligence will occur when AI systems gain “infinite, perfect memory”.Recent advances from the ChatGPT creator, as well as those of its rivals working on large language models (LLMs), have focussed on improving AI’s reasoning abilities.

But speaking on the Big Technology Podcast, the Mr Altman said the development he was most looking forward to was when AI can remember “every detail of your entire life”, and that his company was working towards reaching this point in 2026.“Even if you have the world’s best personal assistant... they can’t remember every word you’ve ever said in your life,” Mr Altman said.

“They can’t have read every document you’ve ever written. They can’t be looking at all your work every day and remembering every little detail. They can’t be a participant in your life to that degree. And no human has infinite, perfect memory.“And AI is definitely gonna be able to do that. We actually talk a lot about this—right now, memory is still very crude, very early.

”His comments come just weeks after he reportedly declared a “code red” situation at his company following the launch of Google’s latest Gemini model.Google described Gemini 3 as a “new era of intelligence” when it released the updated AI app in November, with the model achieving record scores across numerous industry benchmark tests.

Mr Altman played down the urgency of the threat posed by Gemini 3, claiming that it was not unusual for OpenAI to respond decisively to new competition.“I think that it’s good to be paranoid and act quickly when a potential competitive threat emerges,” he said.“The same thing happened to us in the past, it happened earlier this year with DeepSeek...

Gemini 3 has not – or at least has not so far – had the impact we were worried it might, but it did identify some weaknesses in our product offering and strategy and we’re addressing those very quickly.”In order to win the AI race, the OpenAI boss said the strategy is to “make the best models, build the best product around them, and have enough infrastructure to serve it at scale.

”ChatGPT currently has around 800 million users, according to OpenAI, representing roughly 71 per cent of the AI app market share. This compares to 87 per cent this time last year.For comparison, Google’s market share has risen from around 5 per cent to more than 15 per cent, with Gemini recently passing 650 million users.

Analysis

Conflict+

Impact+

Core Event+

Future+

Background+

Related Podcasts

These Fake News Sites Targeting Seniors: 15 Million French People Tricked Each Month

12/26/2025

Battlefield 6’s “no AI” stance is under fire after players spotted what appear to be AI‑generated…

12/26/2025

LLMs’ impact on science: Booming publications, stagnating quality

12/26/2025

View All Podcasts →