April 2025 | Athera's Odyssey: India, AI, and the World
Dissecting the possibilities, realities, and complexities of building AI in India
Odyssey readers,
You don’t need us telling you that AI is reshaping our world. Advances abound at breakneck speeds, governments are gearing up for the next technological arms race, and every tech entrepreneur promises to revolutionise our work and life with AI.
You’ll be forgiven for not keeping up with the AI news cycle — it’s overwhelming, fast-moving, and often lacking in local context. That’s why we’ve put together this issue with three interviews offering grounded, India-focused perspectives from
An entrepreneur, discussing his experience building AI from India
A geopolitics researcher, explaining India’s place in the AI arms race
Two seasoned venture capitalists, breaking down their investment strategies
We hope you enjoy it.
Building AI, from India for the World
Reimagining Customer Service
The idea for Beacon began as a problem. At WINDO, their previous startup, Rakesh Vaddadi and Silus Reddy’s global clients needed 24/7 support. So they created a universal search bar for customers to request any action or data. When queries fell by 80%, they knew they were onto something.
“We realised this was a common pain point, we switched tracks, built our tool in a year, and launched Beacon,” Vaddadi said.
Beacon’s aims to make using workplace software easier. They’ve created an AI assistant for B2B tools, guiding users step-by-step and instantly solving support issues.
For example, if your company buys an HR tool, Beacon’s assistant will help you figure out how to use it. It can walk you through KRA submissions, explain why your reimbursements might be stuck, and automate your recruitment process - all without bothering your company’s HR team or raising a ticket with the software provider.
The way customer support has worked since the dawn of the internet is going to fundamentally change, Vaddadi believes. “So far, it’s been reactive. People raise tickets and reach out to human agents. We’re transforming it by putting help at people’s fingertips.”
Though Beacon positions it as customer support, their vision is grander. “We see ourselves as a customer experience software,” he claims. “Customer support is just the third part of a product’s life cycle - the first two being discoverability/adoption and everyday usage. By consolidating all three, we’re covering the entire customer experience - adoption, daily use, and troubleshooting.”
“This single‑solution approach will make complex user interfaces and multiple tools obsolete. Moving forward, AI agents will turn prompts into outputs so you won’t need to learn softwares. We’ve seen this with clients who start using our tool for one purpose, but soon embrace its cross-functionality.”
One Solution, Many Markets
By consolidating customer experience, Beacon serves multiple product markets, opening them to USD 60 billion worth of worldwide customers. Currently, India and South East Asia account for 70% of their business, and the US the rest.
“Though we expect that ratio to switch eventually, India’s a good market to go from 0-1,” Vaddadi explains. “Clients are patient here, allowing us to iterate. Fewer competitors also make it easier to pitch companies.”
This ease comes with limitations, though.
“The US is definitely a richer market. Indian companies with an ARR (annual recurring revenue) of $10 million are held in high esteem. But one day, the CEO of Infinite Uptime referred to them as a ‘small company’. I thought he was being unnecessarily humble,” he emphasised. “He’d hit an ARR of $13 million building from Pune, after all. But he told me how the US works at a different scale altogether - at that value, you’re just one among many in the American market.”
This scale affects the client pool and expectations. In price-sensitive India, customers of a premium product like Beacon want something multi-purpose. American companies want depth - they’re willing to pay for a single problem solved perfectly.
“More money and clients foster fierce competition. Every college grad is building something,” he remarked. “If the Chief Technology Officer of an Indian firm sees 2-3 product pitches a month, a US one sees about 10.”
“It’s harder to reach the demo stage, but after that conversion is easies. It’s the opposite in India. You can demo easily, but converting at that price point requires sweetening the deal through broader functionality.”
Building from India, Competing with the World
How realistic is it, though, trying to capture the global market building from India? India’s known for its IT services, but will this help in the AI age? Has India’s talent pool matured enough? And can beacon, a company that’s reimagining the very definition of customer service, go beyond the legacy of call centres and BPOs?
Vaddadi certainly thinks so.
“I firmly believe in India, though people try convincing me otherwise. ̨Many told us to set up in the US, that Indians couldn’t afford our product. It’s like they were reading from a 20-year-old textbook,” he insisted. “I’m confident there’s a strong Indian market to go from 0-1. Especially because even legacy companies see the benefit of new tech, you’re not restricted to startups.”
With democratized resources in a global world, India’s got the pay capacity to attract top talent, making it a good base. American companies recognise that, Vaddadi says, building from here while their leadership sits in the US.
“Yes, once I cross the 100 Indian companies in my target, I’ll need to look at the US, he adds. “The Middle East doesn’t have enough companies to be worthwhile, while Europe is too bureaucratic for my liking.”
“When I earn enough in dollars, I’ll set up operations there. For now, I’m happy to work with partners who can help there. The cost of mistakes is also higher there. Once I create a mature product here, I’ll start operations abroad.”
Building Through Hype
As one of many companies riding this wave, Vaddadi is disenchanted with the hype around AI. “People expect it to work like magic,” he explains, “but that’s wrong. ‘Magic’ is otherworldly - you snap your fingers and something happens. I think it’s ‘magical’ - things seem easy to users, but there are complex systems making it happen.”
Stoking this hype are a flood of AI products, some of which he feels make unrealistic promises. “One client first turned us down, saying that existing products they used assured similar functionality,” he said. “Soon they came back to us, realising that was far from reality.”
Many add a layer of AI to workflows without genuine innovation, and he expects these to be subsumed by growing foundational models like ChatGPT and DeepSeek. Photo editing tools are losing relevance in two years now that bigger models have co-opted that functionality.”
Will Beacon be rendered irrelevant, then?
“Foundational models are trained on publicly available data and need intrusive access to sensitive information. Our clients obviously won’t allow this, so we build privately for each of them. Going vertical is how we win.”
But there are other risks that worry Vaddadi, mostly regulatory.
“Regulations are still loose in India and the US, but won’t stay that way. The Saudis already banned AI from the BFSI (Banking, Financial Services, and Insurance) sector. Others will likely follow,” according to him.
“Policymakers might find it hard to separate hype from reality, given the rapid pace of change. They’ll need to regulate based on short-term hype, possibly restricting long-term potential. You already see this with FinTech - a single RBI notice can shut down entire businesses,” he continues.”
“If that happens, people’s belief in new technology is shaken. And ultimately, you’re always selling to people, not organisations. If they believe that AI integrations create uncertainty, it’ll become a much harder sell.”
India’s Place in the AI Arms Race
Understanding the ‘AI Arms Race’
In the two years since ChatGPT’s public release, Artificial Intelligence has evolved from a fascinating tool to the focus of geopolitical conflict. “Governments see a clear link between technology and national power,” says Pranay Kotasthane, Chairperson of the High-Tech Geopolitics Programme at the Takshashila Institution. “That’s why, globally, it’s become the primary domain of both conflict and cooperation.”
Why are governments so heavily invested in this, suddenly? Kotasthane makes sense of it through the frame of ‘creative insecurity’.
“Mark Taylor proposed this in The Politics of Innovation, saying that when the sum total of external threats exceed internal divisions, nation states heavily invest in research and development [R&D] ,” he remarked.
Basically, if a country feels threatened by another, it tries to gain supremacy in competing domains. It explains why, though Denmark and Sweden have similar GDPs per capita, Sweden’s constant threat from Russia caused it to invest more in R&D.
Growing geopolitical tensions in recent years have triggered this creative insecurity, harking back to the Cold War-era conflict between the USA and USSR. This time, of course, China emerged as the USA’s leading competition, with the sudden launch of DeepSeek considered a modern-day ‘Sputnik moment’.
India isn’t immune to creative insecurity, and it reflects in the state’s changing perception of external threats. Kotasthane adds, “[F]or the last 10-15 years, we’ve seen a shift in national perception from Pakistan to China being our biggest threat. With China, we face an adversary that’s economically and technologically stronger, which is why the Indian government’s focusing on innovation, not just the number of military columns.”
Achieving Strategic Independence
Countries are responding to this competition by de-risking their AI supply chains through decoupling - they’re limiting dependence on other countries and preventing their competitors from strengthening.
“This isn’t possible at a macroeconomic level - the world is too interlinked. But the consequences of decoupling aren’t well-understood in technology supply chains, so nation states are decoupling here at the level of materials, machines, humans, and values,” according to Kotasthane.
This is evident in the USA’s Framework for Artificial Intelligence Diffusion (popularly called the AI Diffusion Rules), controlling the export of AI chips manufactured by US companies to limit other countries.
It’s also the perspective underlying a recent paper from a team including Eric Schmidt, a former Google CEO. Similar to the idea of Mutually Assured Destruction from the nuclear age, the paper warns of the spectre of catastrophic errors caused by premature AI deployment. Fundamental to its recommendations is a philosophy that kneecaps other countries’ AI development to maintain US power.
“Though this isn’t official US policy yet, the narrative of maintaining ‘compute governance’ - mechanisms that will allow the US to remotely disable AI chips being used to train workloads that violate certain rules or thresholds - is strengthening,” Kotasthane warns. “This will prevent other countries from using chips for advanced AI development.”
Try as governments might, there are limits to decoupling. Even in semiconductor supply chains - the subject of similar geopolitical tensions - decoupling between the USA and China is restricted to high-end semiconductors; low-end ones are traded freely and exempt from the latest US export controls.
To avert decoupling from a trading partner, countries must develop unique AI capabilities to use as leverage.
Breaking Down the Battleground
In this fast-evolving landscape, different players are trying to build supremacy in their own ways.
“AI technology is based on four elements - infrastructure, data, models, and applications - and different countries are prioritising different parts of the stack from their own perspectives,” Kotasthane notes.
According to the Takshashila Institution’s State of AI Governance 2024 report, the USA is balancing pro-market regulations with its geopolitical concerns, protecting its advantage in innovation. The EU is prioritising transparency, accountability, and individual rights. China’s approach emphasises national security and heavy state control, while India is opting for light-touch regulations and investing in indigenous AI models suited to local use cases.
“Besides governments, there are private companies like OpenAI, NVIDIA, AMD and others which have been covered, so I won’t dwell on them,” he remarks. “But a crucial less-discussed aspect is the open source landscape, beyond the domain of governments and private players.”
Open source AI refers to code for training and running AI models that’s publicly available and free to use. But this isn't just about free access - it enables collaboration, accessibility and innovation outside the purview of governments and big tech.
Kotasthane explains why this is crucial. “This ecosystem helps nations build strategic digital autonomy. If you have access to an open source tool, firms can’t monopolise a market and it;s harder for governments to deny other countries technology. Even Prime Minister Modi highlighted the importance of developing open source AI at the Paris Summit - it’s crucial to reduce our dependence on others.”
India in the World
So where does India stand in the AI race?
“Across the four parts of the AI tech stack, India’s capabilities are strongest in developing applications based on AI models - it’s why Nandan Nilekani spoke of India becoming the AI use case capital of the world,” Kotasthane opined. “There’s also progress on the data front, with the government launching the AI Kosh portal to host datasets that can train AI models.”
Chips aren’t likely to be India’s forte, he feels. “The only semiconductor fabrication plant we’re building will manufacture 28 nanometer chips, which aren’t capable of training and running high-end AI models.”
“To counter our dependence on others for chips, we need to develop alternative strengths. For example, NVIDIA’s chips reign supreme because only they are compatible with CUDA, a software that most AI developers use. But India can support initiatives like the PyTorch Open Source Project and generic open source software frameworks similar to MLX and Triton, which will make the tight coupling between NVIDIA GPUs and CUDA redundant, and allow us to use chips from other companies and countries.”
India’s dependence on other countries was underlined in the AI Diffusion Rules, where the USA didn’t place it among its ‘most trusted allies and partners’ list, fearing leakage of chips to hostile actors like Russia. In doing so, it capped the number of chips that India could import from US companies to 50,000.
This isn’t a significant issue presently. We won’t likely hit this limit soon, and companies can exceed this threshold if granted valid end-user authorizations (essentially, agreeing to increased cybersecurity compliance).
“The real consequence is the need to maintain good relations with the USA, especially in the short-term, because that’s our source of chips,” Kotasthane feels. “Reassuringly, the Indian government is cognizant of this, and the USA recognised India’s complementary capabilities. For example, though India can’t sell the US cutting-edge chips, it’s become a good place to procure trailing-edge chips from at a relatively low-cost.”
“Both countries actively engage each other, issuing joint statements placing technology at the forefront of cooperation. This attitude has persisted across presidential changes, what was earlier the ICET Pro Initiative under the Biden Administration has effectively become the US-India TRUST Initiative under President Trump.”
Charting India’s Priorities
In the long-term, to develop greater AI autonomy, India must strengthen capabilities across the AI stack and establish its comparative advantages that make countries equally dependent on it. The government-run India AI Mission identified the development of a foundational model (a general purpose AI model like GPT 4) as a priority, citing fears that geopolitical tensions could see it cut off from existing models developed abroad.
Kotasthane isn’t completely convinced by this. “There’s no clear market failure I can see here,” he says. “We can always access open source models like DeepSeek and Mistral, so there’s no clear need to build our own. If we did have to build one, it should have been left to the market - government intervention wasn’t necessary.
“The announcement came days after DeepSeek launched, so I’m guessing competitiveness with China is at play, though the government also emphasised that the need for models trained on Indian data and suited to our languages. What India - specifically, Indian companies - should focus on are small language models that are sector specific, ones that can focus on healthcare or education, for example.“
This doesn’t mean the government abdicates all responsibility. On the contrary, there are five things only it can do as an enabler of this ecosystem, he feels.
“First, like I’ve said, it has to maintain good relations with the US to ensure a steady supply of chips and other infrastructure.”
“Second, and related, at the level of discourse it must counter denial regimes - like those advocated in Eric Schmidt’s paper - to ensure that the global AI ecosystem develops well and remains competitive.”
“Third, it must advocate for a thriving open source ecosystem. We’ve progressed rapidly with AI only because things were open so far. Pushing for this will also help solve major bottlenecks like CUDA, and can limit our dependence on NVIDIA.”
“Fourth, it has to catalyse long-term development through academic research. The government’s allowing foreign universities to establish campuses here - maybe it should give them a mandate to explicitly focus on AI research and produce 100 leading PhD researchers working at the cutting edge of AI and ML.”
“Fifth, it has to guard against monopolization within India’s AI ecosystem. When setting up data centres, for example, it can’t onboard just two companies to establish the infrastructure. Protectionism of companies is a key reason why Indian corporate R&D spendings are very low, and why Indian innovation is limited. We highlighted this in a report last year. Incumbents change the rules of the game to prevent international competition; with the Indian market corners, they don’t need to innovate to survive. To stay ahead in AI, the government has to ensure competition across the board.”
On the whole, Kotasthane feels the government’s been quite proactive. “This is true across high tech development - whether it’s semiconductors, the space sector, quantum computing, or AI, the government has announced dedicated missions and, more importantly, put money on the table,” he remarked.
“Compare it to other sectors and you’ll realise the significance. Even though the defence budget hasn’t increased as a percentage of GDP for some time, the government's tech investment has. They’re heading in the right direction, and are even hosting the AI Impact summit next year - a follow-up from Paris - to make India’s presence felt.”
Enabling Success in a Paradigm Change
Catalysing the development of AI in India is inherently different from other deep tech fields, like space and nuclear technology, according to Kotasthane. “In those, the government was the primary customer, capital demands were relatively lower, and you produced fewer products. It’s not so with AI, where private players have outsized roles, there’s a host of use cases across domains, and capital demands are higher.”.
It’s also the opposite of previous technological waves, where tools developed for military use were repurposed for civilian and commercial needs, like the internet. With AI, new technology first enters civilian and commercial domains, after which military uses are identified.
Government leadership in these domains seems capable of adapting to the new paradigm. “Policymakers in charge of this, including the secretary of MeitY [Ministry of Electronics and Information Technology] are qualified engineers,” Kotasthane notes. “The worry is that their capacity might be stretched - they’re leading work on semiconductors, quantum computing, AI, data governance and more, so that could be an issue.”
Encouragingly, Kotasthane feels policymakers have been receptive to external inputs, especially on deep tech development.
“This is evident in liberalisation of the space sector,” he says. “While it had been recommended for some time, it wasn’t an easy sell because ISRO was doing relatively better than its counterparts in other sectors. But the government went ahead, after seeing how the likes of SpaceX were succeeding abroad.”
“In semiconductors, it first hoped that the likes of TSMC would start manufacturing cutting-edge chips here. When it saw that wasn’t going to happen, focus switched to trailing-edge chips, which has been successful. So the government has clearly paid attention to external realities and inputs in similar domains; I’m hoping that continues with AI.”
Investing in a Revolution
Standing at the precipice of a revolution
Amit Somani, Managing Partner at Prime Venture Partners, prophetically declared on LinkedIn recently, “If AI doesn’t scare you, you haven’t experienced it.”
Behind this quip was a trip down memory lane, evoking nostalgia, excitement, fear, and anxiety. “In August 1993, I was among the first few people to download a browser called Mosaic. Before that, we used Gopher, FTP and a few other protocols alien to today’s generation,” he started.
“We computer science graduate students weren’t impressed. We didn’t need a browser, we had tools that worked. This was just someone’s master’s thesis, nothing to be taken seriously,” he continued. “And behold, we know what happened. Not in twenty years, but two. Jim Clark and Marc Andressen launched Netscape and forever changed the internet. That’s what I felt again, this time when using AI.”
Convincing those around him hasn’t been easy. “I’ve been telling my 17-year-old son that AI is more important than anything in school. He thought I was being dramatic. He used ChatGPT and his friends used Claude; none of them saw anything mind-blowing.
So Somani sat him down, doing two projects together on any topic his son chose. Fifteen minutes later, his son sat stunned, head in hands.
“I told him for six months that economics, SAT prep, and college essays were all noise.” Somani continued. “There’s something bigger happening. I kept saying, ‘You have no idea what this is until you play with it, you need to see the power it has!’
Here was Somani’s new Mosaic moment, one that he believed would fundamentally change the next few decades, just as the internet changed the last few.
Rutvik Doshi, General Partner at Athera Venture Partners, had a similar experience. Like Somani, he witnessed the internet revolution in his early 20s. What started with download speeds of 5 bytes per second in IIT has transformed the world around him.
“We’re at that exact moment in time, seeing the tip of the iceberg,” he says. “The first round of AI applications look a lot like the internet of 1995. They’re carrying existing patterns into the AI world. But it’s impossible to comprehend what’s coming in the next 5 years.”
Doshi is describing skeuomorphism - a design principle where new technology resembles its predecessors to build user familiarity. It’s why all email interfaces resemble physical mail; you have ‘inboxes’ for mails received, ‘outboxes’ for mails sent, closed envelope icons to indicate unread mails, and paper clip icons for attachments.
True disruption happens when we break familiar patterns and create use cases native to new technology, and the paradigm-shifting potential of AI has the investor in Doshi excited. “The internet created an unprecedented amount of wealth in the US in the last 15 years. After a lull post the dot com bubble burst, things suddenly took off around 2008 and generational companies grew to have trillion-dollar market caps.”
“Unprecedented,” he repeats, emphasising the extraordinary value the internet created. We’re on the verge of that with AI, and he sees a similar path ahead. “This means wealth, upliftment from poverty, and so much more. We can’t predict how long it will take, or whether there’ll be a boom-and-bust cycle, but it will happen.”
The makings of disruption
Somani was at IBM Research when the dot com revolution happened, was involved in India’s early startup wave up when cloud computing took off, and is now an investor at the dawn of the AI era. He’s convinced what’s coming will be more dramatic than what’s been.
“As an investor, we’re constantly wary of opportunity costs,” he notes. “We will cherry-pick startups that we hope will be large, viable, standalone companies. But it’s not just startups innovating. Incumbents like Google, Meta and Amazon are building at a breathtaking pace. They’ve got capital, distribution, and customers, which gives them power. This makes it harder for startups to be disruptive, and for us to identify disruptive startups.”
But that’s the heart of venture capital, he emphasises. High risk, high return. Though many companies might go to zero, he’s confident that several hundred-baggers will emerge.
It’s not just as an investor that he sees potential, though – he’s excited by its transformative potential as a worker.
“[A]t a recent team offsite, we told everyone that if by January 2026, 30% of their current work wasn’t done by AI, they’d be out of a job. That applies to me as a partner, to my assistant, to our employees. They need to automate 30% of their job with AI and do something meaningful with that time.”
Doshi agrees that finding disruptive startups in the face of innovative incumbents is hard, but notes that this isn’t unique to the AI moment.
“There’s a reason Google built Maps – it had the money, resources, and wherewithal. No startup has that,” he emphasises. “The GPS systems for that were created by governments and defence resources; again, no startup can replicate that.
“Apple created the iPhone and Google, Android. These foundational transformations are made by large companies with deep resources. Entrepreneurs then use these foundations creatively to develop new tools. Uber didn’t invent the smartphone or Google Maps, but they leveraged them to create value.
“You’ll see the same thing with AI. Today’s large companies are building foundational models, defining protocols, etc. In a few years, new entrepreneurs will put these different pieces together to create something transformative. That’s where the opportunity as an investor lies.”
Identifying potential disruptors amongst the hype isn’t easy, though.
“I’ve been investing for 12 years, Amit for 10, and I know I don’t have a crystal ball to predict what’s going to succeed,” Doshi admits. “We know what trends are in place, but we’re two degrees away from the customers, so it’s hard to spot exactly what’s needed.
“What we’ve trained ourselves to do is to spot something that’s bubbling up, to identify entrepreneurs who’ve hit on a gap. It’s a bottom-up approach – I need to constantly talk to people with an open mind and trust that something will emerge.”
Possibilities for India
India rode the dot com boom and Y2K, becoming a global IT hub in the decades since. It benefitted from a large English-speaking populace and low labour costs, but has its tech industry matured enough to enjoy an advantage in the AI age?
Somani certainly thinks so. “We’ve still got a large young and English-speaking population, and also 30% of the world’s developer pool,” he says, “so we’re starting from a strong baseline. And with AI, every mediocre developer can create great tools, further strengthening us. Our IT services companies have also developed deep domain understanding across sectors in this time, and that’s invaluable.”
But where should Indian companies focus their attention to make the best use of their resources? Doshi looks at the top of the AI tech stack – applications – as the right spot.
“We have no advantage in chips, datasets, or foundational models,” he remarks, “not only from a talent perspective but also capital and institutional know-how. It doesn’t make sense to play catch-up there. But we have developers who can build on what’s available and capture value – that’s our advantage. The assumption is that every piece of technology at the bottom of the stack will get cheaper, with most value gained through application development.”
“It’s important to remember that India has a history of leapfrogging in technology,” Somani adds. “Like Nandan Nilekani said we would, we went from data-poor to data-rich in no time. When Rutvik and I were at Google in 2007, we had 2 million broadband users and were stuck there for a while. Now we’ve got almost a billion internet users.
“We’re unique in that way, and we need to find innovative use cases that are India-first and can then be exported to the world.”
While incumbent US tech giants are at the forefront of innovation, these India-first use cases might not come from legacy Indian tech firms like Infosys or Wipro, though. “While American tech giants were product-first companies with heavy R&D,” Doshi observes, “Indian companies were service first. It’s not fair to expect them to be similarly disruptive.”
Identifying investments
Doshi and Somani aren’t paid to just think about artificial intelligence. As venture capitalists, they need to back ideas with investment. That’s easier said than done in this wave of AI hype, though.
“When technology becomes accessible, you see a Cambrian explosion of ideas.” Somani says, “There are maybe 10,000 companies creating chatbots by adding a thin veneer on top of a ChatGPT or Claude without creating value. What stops a big incumbent from doing the same and eating them up?”
He’s wary of slick demos – calling them the ‘kiss of death’ – and searches for deep market insights, nuanced understanding of customer behaviours, and unique problems waiting to be solved (and paid for). He’s looking for entrepreneurs that can bring laggard industries to the forefront of technology.
“Think of manufacturing.” he says, “Earlier, quality control meant sampling some units and extrapolating that quality to the batch. But can you use computer vision and AI to scan an entire shipment of three million units? That’s where the value proposition is.
“The good thing is that every CEO, even for big incumbents, is being asked what they’re doing about AI. They know they need to adapt to this new technology. So the pitch has become easier, and adoption should scale fast.”
And what do the people funding the funders think? What are the priorities of the limited partners backing VC firms when approaching AI?
“There’s no clear mandate from them. They’re curious, excited, and trust our judgement,” Somani said.
“That’s true, they won’t dictate what to do,” Doshi adds, “but if AI becomes big and you haven’t been part of the story, then they’ll question you. If your investment bombs, they’ll check whether you did your due diligence, were financially prudent, and have diversified enough.”
Their funding is just one ingredient that makes an ecosystem, though. They’re clear that India has the talent pool, and Somani’s found the government ‘supremely enthusiastic’. What India lacks, in his opinion, is deep investment in academic research, a hallmark of the USA and China.
“We’ve got talented professors,” he mentions, “many of whom likely got their own degrees in the USA. But with your talent pool gravitating towards startups instead of spending 5 or 6 years tinkering in a lab, there’s not enough deep research going on. Incentivising this should be a policy priority to round out our ecosystem.”
Pathways
In this month’s pathways, we’re sharing recommendations to round out this AI deep dive.
First, an interview of NVIDIA CEO Jensen Huang hosted by Cleo Abrams. NVIDIA started out manufacturing GPUs - graphic processing units - to improve video games. But it’s become central to AI because these GPUs are extremely effective at parallel computations required to train deep learning models.
Their conversation exquisitely covers the past, present, and future of AI in just an hour, and it’s well worth your time.
Second, to keep up with the latest on AI, follow the Dwarkesh Podcast, the show hosted by the 24-year-old wunderkind who’s become an authority on all things AI. The podcast’s archive hosts extremely insightful conversation with global authorities like Satya Nadella and Mark Zuckerberg (linked below), and is peppered with episodes unrelated to AI but that will deepen your understand of the forces shaping history, like the fall of Rome, the 20th century story of oil, and the history of human genetics.
Last, you can’t understand AI or the chips that enable it without understanding semiconductors - the materials (generally silicon) these chips are made of. So we’re recommending When the Chips are Down: A Deep Dive into a Global Crisis by Pranay Kotasthane (interviewed above) and Abhiram Manchi.
The book unpacks semiconductor supply chains, explains how their stability is threatened by Taiwan’s precarious relations with China, and answers why India was looking to establish its own semiconductor manufacturing facilities.
If you enjoyed reading this edition of Athera’s Odyssey, do reach out to us at contact@atheravp.com, whether you’re a founder building in AI or any other tech-first sector!