Intelligence Explosion: Part II
National AI strategy and Greece
This is Startup Pirate #109, a newsletter about technology, entrepreneurship, and startups every two weeks. Made in Greece. If you haven’t subscribed, join 6,540 readers by clicking below:
Intelligence Explosion: Part II
Everyone is now talking about AI, but very few have the situational awareness after finding themselves in the epicentre of AI advances of the past few years. An early employee at two of the leading AI labs today (OpenAI and Cohere), Aris Konstantinidis, joined us for a two-part series. After sharing stories from the launch of ChatGPT, discussing the technological breakthroughs that led to GPT-4, open vs closed source models, and predictions about when we reach AGI in Part I, today, we present Part II.
Let’s delve into AI safety, building a national AI strategy, and Greece.
Ari, I want to switch gears in this second part… What’s your take on AI safety?
AK: It's a really important discussion, and I'm glad it's taking place now that model capabilities are still fairly limited. You don't want to start thinking about safety when the stakes are super high. This is an interdisciplinary field that focuses on preventing accidents, misuse, or other harmful consequences arising from AI. Contrary to what some people think, I believe that the majority of AI labs do care about AI safety. It's also great to see government initiatives and new regulations both in the EU and the US. The public and private sectors should be working side by side on that front.
I'm personally not in the AI doomers camp that suggests AI will destroy humanity, nor do I support halting scientific research, which was proposed a couple of years ago. If we do that, we risk slowing progress towards many beneficial AI applications that can tremendously improve our lives. At the same time, you give a head start to potential bad actors who can build powerful technology without any guardrails. So, the well-intentioned actors must continue to make rapid progress, with safety as a top priority.
In one of our previous chats, you mentioned you had worked with governments on their AI strategies, so I want to double-click on that. What are the different components which can form a national AI strategy? Let's take Greece as an example.
AK: There's growing interest among nation states in building AI capabilities such as data, compute, LLMs, talent, etc. While at Cohere, I had a chance to engage with some of these efforts.
Canada, for example, was the first country to formalise a national AI strategy and has been a pioneer in shaping AI research and policy. They're now looking into making targeted investments, such as building data centres for AI and subsidising access to compute, to support the local AI ecosystem.
The starting point is often setting up AI research institutions or subsidising research, entrepreneurship, and education. Some countries take it a step further and focus on building custom national LLMs—large language models that the country's citizens, enterprises, and public agencies can use and are trained on data specific to their language and cultural norms. Most of the AI models out there are not great in languages other than English. So, as it currently stands, their use and potential impact are limited. For example, some countries in Asia and the Middle East are actively working on building custom LLMs and distributing them across the public and private sectors.
It's in Greece's best interest to actively pursue partnerships with frontier research labs to ensure that the world's best AI models are proficient in Greek and knowledgeable of Greek heritage and cultural norms. This could involve providing access to public and private datasets or setting up a workforce to generate human feedback on model outputs. Starting from scratch and competing against model builders like OpenAI, Anthropic and Meta to build better models doesn't make sense. AI models well-versed in Greek would unlock use cases in almost every vertical, thereby increasing automation and productivity in areas where the country absolutely needs it, such as the public sector.
A second critical pillar is sovereign compute, supported by investments in data centres and affordable, renewable energy. Data centres support AI training, but the future will see even greater demand for AI inference. Locally operated data centres enhance national independence and reduce the cost and latency of running AI models. Many governments are already acting on this. Right now, there is a global surge in data centre investments, and securing permits in regions with robust networking infrastructure and access to cheap power—whether from renewables or nuclear energy—is increasingly competitive. Greece needs to encourage and promote such investments and think about potential improvements needed in the power grid and network infrastructure. We must be strategic and proactive about this because building infrastructure takes time. On the flip side, these efforts will create many new jobs.
Furthermore, Greece needs to tap into its talent pool. You have pointed this out before… There's a high density of world-class Greek AI talent in industry and academia, with some holding key positions at leading AI labs. It's obvious that we should be engaging this diaspora, using their expertise to support the Greek government and enterprises in developing AI capabilities. At the same time, the government should offer resources to fuel research and entrepreneurial ventures. It's a fantastic opportunity to position ourselves favourably in the race for AI adoption and supercharge the economy.
Equally important is educating people and preparing them for the change that's coming. We can do this by updating the curriculum in schools and universities and upskilling the current workforce. AI will reshape industries at an unprecedented pace, and individuals must be equipped to understand how their roles might evolve or how AI can enhance their productivity. The government must proactively address questions about preparing citizens for this transition: How can society adapt to a future where most human tasks are automated? How can businesses and individuals position themselves to thrive in this new landscape? We must start thinking about these pressing issues today to ensure a smooth and inclusive transition.
Which countries are doing well today in forming a national AI strategy?
AK: Here's the 10,000-foot view. The US leads global AI research, has attracted some of the top AI talent in the world, and most of the major AI labs are headquartered there. The US has a longstanding reputation for prioritising innovation over regulation, which helps the tech ecosystem flourish. There's now a strong push at the federal and state levels to accelerate AI adoption, with legislative efforts and significant investments, including chip production. I'm confident the US will continue to lead in AI development and deployment.
Europe takes a different approach, focusing on regulation, ethics, and safety. The European AI Act exemplifies this. Countries like France and the UK have been particularly strong in AI research. France, for instance, is actively supporting companies like Mistral AI in establishing a European counterpart to OpenAI while investing in data centres and other initiatives. But Europe clearly needs to do more. We need to attract capital, drive investments in infrastructure, provide incentives for AI talent to return, and create a friendlier environment for entrepreneurship and innovation.
Lastly, China has a clear ambition to lead in AI by 2030, backed by massive government support. However, challenges like trade embargoes make it difficult for them to access advanced chips and resources, and keep up with the US.
To wrap up, I'm going to use a quote by your former CEO, Sam Altman. Is "compute going to be the currency of the future"?
AK: He did say that in a podcast with Lex Friedman, right? Today, money is the currency we all use. Everyone has a budget, and we decide how to spend it on goods and services. But the idea behind what Sam and others, who believe we'll be able to create AGI and Superintelligence, are saying is that, in the future, AI will drive most of the global GDP. Almost every service we use will have AI models behind it. There will be abundance, and people won't need to work as they do today because AI will handle most of the heavy lifting. We'll be able to spend more time with our loved ones or do the things we love. We'll need to find meaning in things other than work. Maybe it's art, philosophy, space travel, or human interaction—who knows?
In that world, however, it is possible that compute will not be truly abundant. Every person will be allocated part of the world's / each country's available compute to perform their daily tasks using AI: plan a family trip, find personalised treatment for their back pain, ask their self-driving car to take them to the movies, etc. As a result, countries with access to cheap power and plenty of compute will be able to provide their citizens with a higher quality of life. It sounds futuristic, but I think it's not entirely crazy to consider compute the main limiting factor in a future like that.
Thank you Ari, had a blast!
AK: Appreciate it, Alex.
Jobs
Check out job openings here from startups hiring in Greece.
News
Plum (fintech) raised €5m from Eurobank.
PeopleGoal (HR tech) was acquired.
Promise (AI-driven filmmaking) secured funding from a16z.
KoRo (D2C food brand) raised €35m.
Amazon invests in first utility-scale wind energy projects in Greece.
Resources
Panos Papadopoulos, Partner at Marathon Venture Capital, on the Greek defense tech industry.
Revolutionising sensors for robotics with Klajd Lika, CEO & founder at Bota Systems.
AI agents thinking fast and slow by Konstantina Christakopoulou, Tech Lead at Google Deepmind.
Creating dexterous robots with Minas Liarokapis, CTO & co-founder at Acumino.
Getting into crypto and building Sui with Kostas Chalkias, co-founder & Chief Cryptographer at Mysten Labs.
Quantum technology for financial transactions with Yianni Gamvros & Iordanis Kerenidis, co-founders at Quantum Signals.
From starting up to getting acquired by Xero with Vangelis Kyriazis, co-founder and CEO at Syft Analytics.
Performance Marketing fundamentals that are not discussed enough by Michael Lorenzos, Growth Marketing Consultant.
A report on Greece’s AI transition from the National AI Advisory Committee.
Events
Join us at Open Coffee Athens #120 on Dec 13! RSVP here
“UX Greece welcomes Jesse James Garrett” by Thessaloniki UX Meetup on Dec 4
“START your night UP” on Dec 4
“Socratic Seminar #12” by BitDevs Athens on Dec 4
“Angular Athens 24th Meetup” on Dec 5
“Greeking Out in New York” by Endeavor Greece on Dec 5
“Fintech Talks goes Digital Currencies” by Greek Fintech Hub on Dec 5
That’s all for this week. Tap the heart ❤️ below if you liked this piece—it helps me understand which themes you like best and what I should do more.
Thank you so much for reading,
Alex
Loved reading thru this! Any chance for a part 3?