Anthropic's co-founder and chief scientist Jared Kaplan discusses AI's rapid evolution, the shorter-than-expected timeline to human-level AI, and how Claude's "thinking time" feature represents a new frontier in AI reasoning capabilities.

In this episode you'll hear:

• Why Jared believes human-level AI is now likely to arrive in 2-3 years instead of by 2030
• How AI models are developing the ability to handle increasingly complex tasks that would take humans hours or days
• The importance of constitutional AI and interpretability research as essential guardrails for increasingly powerful systems

Our new show 

This was originally recorded for "Friday with Azeem Azhar", a new show that takes place every Friday at 9am PT and 12pm ET on Exponential View. You can tune in through my Substack linked below. The format is experimental and we'd love your feedback, so feel free to comment or email your thoughts to our team at [email protected].

Timestamps:

(00:00) Episode trailer

(01:27) Jared's updated prediction for reaching human-level intelligence

(08:12) What will limit scaling laws?

(11:13) How long will we wait between model generations?

(16:27) Why test-time scaling is a big deal

(21:59) There’s no reason why DeepSeek can’t be competitive algorithmically

(25:31) Has Anthropic changed their approach to safety vs speed?

(30:08) Managing the paradoxes of AI progress

(32:21) Can interpretability and monitoring really keep AI safe?

(39:43) Are model incentives misaligned with public interests?

(42:36) How should we prepare for electricity-level impact?

(51:15) What Jared is most excited about in the next 12 months

Jared's links:

• Anthropic: https://www.anthropic.com/

Azeem's links: 

• Substack: https://www.exponentialview.co/
• Website: https://www.azeemazhar.com/
• LinkedIn: https://www.linkedin.com/in/azhar
• Twitter/X: https://x.com/azeem

Produced by supermix.io

Kevin Kelly is a co-founder of Wired Magazine and a renowned author and futurist.  Decades ago, Kevin predicted much of today's technological and cultural landscape. In this discussion, he presents his new bold vision for what’s coming next: The Handoff to Bots.

In this episode, you’ll hear:

• Why declining populations will radically reshape economies
• What a bot-to-bot economy could look and feel like
• Why people of the future might be paid to read emails
• How AI could help humanity find deeper purpose
• Why this future might be closer than you think

Kevin’s links:

Website/blog: https://kk.org/

Twitter/X: https://x.com/kevin2kelly

Instagram: / kevin2kelly  

Azeem's links:

Substack: https://www.exponentialview.co/

Website: https://www.azeemazhar.com/

LinkedIn: https://www.linkedin.com/in/azhar?ori...

Twitter/X: https://x.com/azeem

Timestamps:

(00:00) Intro

(02:17) The baby black hole behind Kevin's theory

(10:49) Kevin's thesis: The handoff to bots

(15:05) This world is closer than we think

(19:32) The role of humans in this new world

(21:23) Could monopoly influence pose a problem?

(28:33) The nature of “struggle” in this new world

(32:42) Could we see countries competing for population?

(36:06) How a scarcity of humans might change what we value

(42:30) What would 1994 Kevin think of 2025 Kevin's blog?

Production:

Production by supermix.io

Kai-Fu Lee joins me to discuss AI in 2025. Kai-Fu is a storied AI researcher, investor, inventor and entrepreneur based in Taiwan. As one of the leading AI experts based in Asia, I wanted to get his take on this particular market.

Key insights:

• Kai-Fu noted that unlike the singular “ChatGPT moment” that stunned Western audiences, the Chinese market encountered generative AI in a more “incremental and distributed” fashion.
• A particularly fascinating shift is how Chinese enterprises are adopting generative AI. Without the entrenched SaaS layers common in the US, Chinese companies are “rolling their own” solutions. This deep integration might be tougher and messier, but it encourages thorough, domain-specific implementations.
• We reflected on a structural shift in how we think about productivity software. With AI “conceptualizing” the document and the user providing strategic nudges, it’s akin to reversing the traditional creative process.
• We’re moving from a training-centric world to an inference-centric one. Models need to be cheaper, faster and less resource-intensive to run, not just to train. For instance, his team at ZeroOne.ai managed to train a top-tier model on “just” 2,000 H100 GPUs and bring inference costs down to 10 cents per million tokens—a fraction of GPT-4’s early costs.
• In 2025, Kai-Fu predicts, we’ll see fewer “demos” and more “AI-first” applications deploying text, image and video generation tools into real-world workflows.

Connect with us:

• Exponential View

Nathan Benaich, Founder and General Partner of Air Street Capital, joins me to discuss AI in 2025. From runaway consumer adoption to evolving enterprise moats, from still-elusive AI-driven drug breakthroughs to the renewed vigour in robotics, several core themes stood out.

1. Frontier models & AI at scale

In 2024, we witnessed the astonishing growth of frontier models and their deployment on a massive scale. OpenAI’s GPT-4 and GPT-4 o1, Anthropic’s Claude and Google’s Gemini have all demonstrated that being “at the frontier” is increasingly the price of admission.

2. Consumers, voice and infinite worlds

On the consumer side, we have reason to believe 2025 will be the year of AI-enabled workflows that feel truly natural. Voice, multimodality and integration into daily routines—like transcribing my morning thoughts during a commute—are becoming routine.

3. Accelerating science & drug discovery

While AI accelerates lab automation and data analysis—improving reproducibility and speeding up processes—the promised “AI-designed blockbuster drug” is still in the pipeline. Clinical timelines and regulatory hurdles do not compress easily.

4. Geopolitics, funding and the sovereign question

As training costs skyrocket and models require unimaginable scale, questions mount… Who funds these massive compute requirements? Will nation-states view these labs as strategic assets, akin to telecoms or chipmakers?

5. From explosive capability gains to refined utility

We’ve grown numb to what was once astonishing—perfect speech synthesis, infinite text generation, zero-shot coding. The capabilities of models now surpass human levels in many benchmarks. The next major shifts may be subtler, or simply less obviously spectacular.

Connect with us:

• Exponential View
• Nathan Benaich
   

Dylan Patel, founder of SemiAnalysis and one of my go-to experts on semiconductors and data center infrastructure joins me to discuss AI in 2025. Several key themes emerged about where AI might be headed in 2025:

1/ Big Tech’s accelerating CapEx and market adjustments
The hyperscalers are racing ahead in capital expenditure, with Microsoft’s annual outlay likely to surpass $80 billion (up from around $15 billion just five years ago). By mid-decade, total annual investments in AI-driven data centers could climb from around $150–200 billion today to $400–500 billion. While these expansions power more advanced models and services, such rapid spending raises questions for investors. Are shareholders ready for ongoing, multi-fold increases in data center build-outs?

2/ The competitive landscape and new infrastructure players
The expected explosion in AI workloads is drawing in a wave of new specialized GPU cloud providers—names like CoreWeave, Niveus, Crusoe—each gunning to become the next vital utility layer of AI compute. Unlike the hyperscalers, these players tap different pools of capital, including real-estate-like finance and private credit, enabling them to ramp up aggressively. This dynamic threatens the established order and could squeeze margins as competition heats up. The market is starting to understand that.

3/ The semiconductor supply chain isn’t the only bottleneck
We often talk about GPU shortages, but the real sticking point is broader infrastructural complexity. Yes, Nvidia and TSMC can ramp up chip supply. But even if you have enough high-end silicon, you still need power infrastructure and grid connectivity. Building multi-gigawatt data centers in the US—each the size of a utility-scale power plant—is now firmly on the agenda. In some states, data centers already consume 30% of the grid’s electricity. By 2027, AI data centers alone could account for 10% or more of total US electricity consumption, straining America’s aging infrastructure.

4/ Commoditization of models and margin pressure
A year ago, advanced language models were scarce and expensive. Today, open-source variants like Llama 3.1 are driving commoditization at speed, slicing away the profit margins of plain-vanilla model-serving. If your model doesn’t outperform the best open source, you’re forced to compete on price—and that’s a race to the bottom. Currently, only a handful of players (OpenAI and Anthropic among them) enjoy meaningful margins. As models proliferate, value will increasingly flow to those offering distinctive tools, integrating closely into enterprise workflows and locking in switching costs.

5/ Into 2025: exponential curves and new market norms
Despite these challenges—soaring costs, stalled infrastructure build-outs, margin erosion—Dylan is confident that exponential scaling will continue. The sector’s appetite for GPUs, specialized chips and next-gen data centers appears insatiable. We could easily see record-breaking fundraising rounds north of $10 billion for private AI ventures—funded by sovereign wealth funds and other capital pools that have barely scratched the surface of their capacity to invest in AI infrastructure. There’s also a very tangible productivity angle. AI coding assistants continue to reduce the cost of software development. Some software companies could be looking at 20–30% staff reductions in these technical teams as high-level coding becomes automated. This shift, still in its early days, will have profound downstream effects on the entire software ecosystem.

Find us:

• Exponential View
• SemiAnalysis

As we race towards a future powered by AI and data centres, how will the insatiable demand for energy impact the environment? With the richest companies ploughing billions into energy generation, might there be some unexpected upsides for the climate transition? And can exponential technologies address the climate crisis on a finite planet? 

Cleaning Up host Michael Liebreich sits down with Azeem Azhar, founder of Exponential View, to explore the complex relationship between exponential growth, climate change, and the societal implications of transformative technologies. Michael and Azeem delve into the promises and pitfalls of a future shaped by the rapid advancements in renewable energy, battery storage, and artificial intelligence. 

This podcast was originally published on Cleaning Up. 

Artificial Intelligence is on every business leader’s agenda. How do we make sense of the fast-moving new developments in AI over the past year? Azeem Azhar returns to bring clarity to leaders who face a complicated information landscape.

This week, Azeem speaks with Richard Socher, CEO and founder of You.com, an AI chatbot search engine at the forefront of truthful and verifiable AI. They explore approaches to building AI systems that are both truthful and verifiable. The conversation sheds light on the critical breakthroughs in AI, the technical challenges of ensuring AI’s reliability, and Socher’s vision for the future of search.

They also discuss:

@azeem

Further resources:

As 2024 begins, leaders are facing increasing uncertainty and a host of difficult decisions. Azeem Azhar returns to bring clarity amid a complicated information landscape, with his analysis of 12 core themes that will shape the year ahead, including AI adoption, geopolitics, decentralization, the energy transition, and more.

The discussion specifically touches on:

@azeem

Further resources:

Artificial Intelligence is on every business leader’s agenda. How do we make sense of the fast-moving new developments in AI over the past year? Azeem Azhar returns to bring clarity to leaders who face a complicated information landscape.

Generative AI has a lot to offer health care professionals and medical scientists. This week, Azeem speaks with renowned cardiologist, scientist, and author Eric Topol about the change he’s observed among his colleagues in the last two years, as generative AI developments have accelerated in medicine.

They discuss:

@azeem

Further resources:

Artificial Intelligence is on every business leader’s agenda. How do we make sense of the fast-moving new developments in AI over the past year? Azeem Azhar returns to bring clarity to leaders who face a complicated information landscape.

This week, Azeem joins Sasha Luccioni, an AI researcher and climate lead at Hugging Face, to shed light on the environmental footprint and other immediate impacts of AI, and how they compare to more long-term challenges.

They cover:

@azeem

Further resources: