#59 Assessing India's Quantum Plans, AI Talent Retention, and EU's Scrutiny of Emerging Tech
Assessing shortfalls in the development of emerging technology
Cyberpolitik 1: Enabling Quantum Ecosystems - Thoughts on C-DAC's Quantum Chip Design Proposal
— Rijesh Panicker
The Centre for Distributed and Advanced Computing (C-DAC) has published an Expression of Interest (EOI) proposal seeking Collaborative Development Partners (CDP) to design and manufacture general quantum chips scaling between 10-50 qubits. The requirement specifies that the chip must be a superconducting quantum processor, using any of the most common underlying qubits types, such as transmon, xmon, and fluxonium qubits. The long-term vision for these quantum processing units (QPU) is to integrate them as part of a hybrid computing system, with the QPU providing accelerated compute to a supercomputer system.
The engagement is set to run for three years initially, with an option to extend for two more years. The proposal is open to firms registered in India with at least a two-year record of operations. To be selected as a CDP, firms will need either experience in quantum chip design and development or patents in quantum chip development. Selected firms need to have the capabilities to design and manufacture these quantum chips or have partnerships in place for key steps such as design, development, fabrication, cryogenic testing and packaging.
Based on a quick read of the EOI, here are a few thoughts, ordered purely by when they occurred to me, on changes one might consider for future proposals
First, it's important to avoid conflicting goals in a single proposal. If the outcome is to test QPU integration with supercomputers, it's easier to go with a known player like IBM or Quantinuum, who could quickly supply the necessary quantum infrastructure. Waiting three years to build a quantum chip to build hybrid quantum supercomputers seems inefficient.
On the other hand, if the objective is to kickstart quantum chip design and manufacturing in India, perhaps a broader design challenge competition, where startups could have made multiple prototyping proposals across technologies like superconducting qubits, ion trap, neutral atoms, cold atoms, photonics would have provided more interesting outcomes and learnings, both on the depth of the market in India and ability to garner partnerships.
Second, the proposal is open only to companies registered in India. A quick search on Google throws up Indian startups focused on post-quantum encryption or quantum communication and quantum key distribution technologies. Research on quantum chip design is mainly confined to academic research labs at IISER, IISc and similar institutions. Of course, large firms in the space, like IBM and Microsoft, with operations in India, may choose to respond, but that, in my opinion, defeats the purpose.
A more interesting question here is why, with the industry at a nascent stage in India, should we restrict participation to Indian firms alone. Why not allow participation by firms globally, especially startups from countries with whom we seek to collaborate in quantum tech? Allowing global firms to participate, perhaps in consortia with Indian partners, will benefit the overall ecosystem in India.
Third, how should we treat intellectual property (IP) generated from these projects? C-DAC is proposing that all new IP created from the work will be owned by it in perpetuity. This, in my opinion, makes the proposal a lot less interesting and will disincentivise firms from participating. Since our overarching aim is to promote a deep ecosystem in India, we should consider making IP sharing a default for publicly funded projects. An alternative would be to create an IP/Patent Pool for quantum research in India, which allows contributions from academia, government-funded projects and industry to be broadly and reasonably accessible and provide monetisation opportunities for the research.
Given India's current industry depth in quantum computing, state-led and funded projects are a necessary first step. Being generous with how we share our intellectual property rights with early-stage participants and actively inviting and incentivising firms globally to participate in our projects are two simple ways in which we could enable the creation of a sustainable ecosystem of firms and partnerships for India to leverage.
Cyberpolitik 2: Bridging the AI Talent Gap
— Bharath Reddy
**The following is an excerpt from an upcoming Takshashila discussion document on AI Governance.**
Developing cutting-edge AI systems demands substantial resources in terms of data, computing power, and technical expertise. The industry can mobilise these resources better than academia, and the recent breakthroughs in AI research indicate this - Google's paper on transformer models and Microsoft's paper on Low-Rank Adoption. This trend is underscored by the Stanford AI Index report, which reveals that in 2022, industry entities produced 32 significant machine learning models, in stark contrast to academia, which contributed only three.
Data and computing are vital components in the training of AI systems. We delve deeper into the challenges and obstacles related to these domains in the sections focused on data and computing.
In their paper discussing India's AI potential, Chahal et al. highlight some key observations. India produces nearly twice as many master's level engineering graduates as the United States, second only to China in this regard. However, India significantly lags behind the United States in producing PhDs, with less than one-third of the number. The shortcomings within India's higher education sector limit its ability to offer extensive training for a highly skilled AI workforce. Consequently, many Indian students opt to pursue PhD programs in foreign countries. The ASPI critical technology tracker clearly shows the brain drain of Indian AI talent to other countries, notably the United States.
Indian researchers publish AI-related papers at a prolific rate, trailing only behind the United States and China. However, when ranked by H-index, which measures both the productivity and citation impact of publications, India descends to the 16th position, indicating that the quality of these publications falls short of expectations.
India's research and development (R&D) investment is a mere 0.64% of its GDP, significantly lower than other nations. Of this amount, approximately 37% comes from the private sector. In contrast, China allocates 2.4% of its GDP to R&D, and most developed countries devote more than 2% of their GDP to research and development. This implies that India's scientific research in AI is likely underfunded compared to many other countries.
On a positive note, according to a report from Bain & Company, India stands out as a significant global source of talent in data and AI skills. It contributes 16% of the world's AI talent pool, ranking it among the top three talent markets worldwide. However, not all AI talent is equal. Top-tier AI researchers are involved in creating intellectual property or designing and training AI algorithms, which are activities at the top of the value chain. A study by MacroPolo, a US-based think tank, finds that over 80% of India's top-tier AI researchers move out of the country.
Consequently, while the specialised skills required for research and training AI models may be in shorter supply, India still holds a substantial advantage in engineering, which is likely also an advantage in developing applications based on AI models.
Cyberpolitik News: EU Recommends Risk Assessment for Critical Technologies
— Saurabh Todi
The EU recently recommended a comprehensive risk assessment of the critical technology areas for its economic security to its member states. The EU has broadly identified four areas for which this risk assessment should be carried out:
Advanced semiconductor technologies (microelectronics, photonics, high-frequency chips, semiconductor manufacturing equipment)
Artificial Intelligence technologies (high-performance computing, cloud and edge computing, data analytics, computer vision, language processing, object recognition)
Quantum technologies (quantum computing, quantum cryptography, quantum communications, quantum sensing and radar)
Biotechnologies (techniques of genetic modification, new genomic techniques, gene drive, synthetic biology).
According to the EU, these technologies were chosen because of their enabling and transformative nature, their potential and relevance for increasing efficiency and performance, the potential for dual use, especially in the military domain and the risk of the technology being used to violate human rights.
This is a significant step as it indicates that the EU is thinking deeply and comprehensively about these technologies' geopolitical and geoeconomic implications. One should expect that the outcome of such risk assessments would guide the trade and investment policies of member states and the EU as a whole. Moreover, comparing the EU and the US risk assessments would be interesting. Any significant divergence between the two, especially in how they see China and their dependence on it, would indicate the cleavages that China could exploit and weaken the joint response to the China challenge.
As India pushes for deeper engagement with the EU, especially with the ongoing trade agreement negotiations, working with the EU to increase bilateral synergy and address identified risks would be worth pursuing.
What We're Reading (or Listening to)
[Blog] India-Australia Defence Partnership: Major Drivers and Elements, by Bharat Sharma and Josiah W. Neal
[Podcast] Have the 1-Year old US Chip Sanctions Failed [All Things Policy] ft. Aditya Ramanathan and Satya Sahu
[Opinion] Championing Pluralism Globally will make India a ‘Vishwaguru’, by Nitin Pai