#48 Accessible AI, China's Science Policy, Artemis Unstrung, and India's 'Right to Repair' Endeavour.
AI: Is there no secret sauce? , China’s Emerging Policy Focus on Science and Tech Advancement, An IAEA type regulation for Artificial Intelligence? , Electronics Repair Services Outsourcing Pilot.
Please note that owing to technical difficulties, the audio transcript of the text will be uploaded with a short delay. We apologise for the inconvenience.
Cyberpolitik 1 : Is there no secret sauce?
— Bharath Reddy
Do you remember reading just a few months ago that it costs OpenAI hundreds of thousands of dollars daily to keep chatGPT running? The costs of training and running a large language model (LLM) were prohibitively expensive, and only a few big tech companies could afford to do it.
That understanding has been turned upside down in the blink of an eye. What happened?
Meta open-sourced their LLM codenamed LLaMA for research in Feb 2023, unleashing a flurry of innovation. Models comparable in performance to the best LLMs can now be trained for a few hundred dollars and run on consumer hardware such as a Macbook or phone.
The graph below shows how different derivatives of LLaMA perform when compared against Google’s Bard and OpenAI’s ChatGPT. This brilliant essay points out that Google or OpenAI no longer have a moat and that people can’t tell the difference between the open-source models and ChatGPT.
Source: Semianalysis.
What does this mean?
AI tools do not need to be one big monolithic app or service, and computing is no longer a significant barrier. Smaller iterative models that build on top of each other and lead to innovative use cases that are not conceivable with larger models. Some call this the Stable Diffusion moment for LLMs. For countries such as India, which have an advantage in talent but lag in infrastructure and capital, access to such LLMs could be especially beneficial. Transparency and inclusivity of open-source models also help foster trust and broaden access to technology.
But with all this excitement come some pitfalls as well. Ease of access to LLMs with no guardrails could also spread misinformation, hateful content, and other online harms. Will we rely on better AI to control the spread of harm? Will AI empower people to be more creative and productive, or will it automate and make a lot of existing jobs redundant?
As the dust settles, some of these questions will be answered. But this is a space we need to watch.
Matsyanyaaya : China’s Emerging Policy Focus on Science and Tech Advancement
— Anushka Saxena
Between May 25-30, 2023, China held the seventeenth edition of its biggest annual technological fete, the Zhongguancun Forum, holding the theme 'Open Cooperation for a Shared Future.' The emphasis of this year's Forum was presenting China's vision for cooperation in the 'technological revolution' and industrial transformation through continued innovation. And in his congratulatory letter to the Forum, Chinese President Xi Jinping furthered this vision, saying:
"The world's major changes unseen in a century are accelerating, and the recovery of the world economy is facing severe challenges. Innovative development and win-win cooperation are the general trend. China is committed to a win-win strategy of opening up and is willing to work with other countries to promote scientific and technological innovation to bring more benefits to people of all countries."
In the past few months, commitment to scientific and technological advancement has become a cornerstone of Chinese policymaking, and it is in China's interests to urge cooperation with other countries, given that today, geopolitical tensions are contributing to increasing diversification and competition in critical technology-related supply chains.
Among the breadth of governmental reforms initiated earlier this year at the 'Two Sessions,' the annual meeting of China's two legislative bodies, the National People's Congress (NPC) and the Chinese People's Political Consultative Conference (CPPCC), the reforms adopted in science and technology were especially significant. For example, the Party and State Council Institutional Reform Plan adopted on March 16 directed the creation of a new Central Science and Technology Commission under the ambit of the Communist Party of China (CPC), and its role, as discussed by policymakers, is "to strengthen the centralized and unified leadership of the Party Central Committee on science and technology work, and to coordinate the promotion of the construction of the national innovation system and the reform of the science and technology system." Under the CSTC is the expanded Ministry of Science and Technology, which is meant to act as an administrative body fostering the CPC's vision for technological advancement.
More recently, on June 2, 2023, the National People's Congress Standing Committee (NPCSC) created an 'Enforcement and Inspection Team' for the National Science and Technology Progress Law. It is reported that the law enforcement inspection team will focus on eight aspects of inspection:
● The progress of the promotion and implementation of the Science and Technology Progress Law (1993) by various local departments and governments;
● The strengthening and storage of basic research-related situations;
● The strengthening of applied research and the transformation of achievements, and the promotion of independent control of key core technologies;
● Strengthening the construction of the dominant position of scientific and technological innovation in enterprises;
● Improving the system and mechanism of scientific research institutions, establishing and improving the scientific and technological research and development system;
● Improving the situation of training, management, and incentives for scientific and technological personnel;
● Implementation of regional technological innovation; Support and guarantee for scientific and technological activities; and
● Implementation of supervisory measures.
Why is this policy focus essential for the international community to acknowledge?
The intensifying US-China rivalry in high technology is manifesting in debates surrounding every piece of globally critical tech - 5G, Artificial Intelligence, Quantum, semiconductor chips, and Advanced Computing. China is many years behind the US and its allies in the production of sophisticated and integrated products, such as in aviation and semiconductors. However, it has gained leaps and bounds in 5G (primarily through massive government support for Huawei), Electric Vehicles, and other clean energy tech like hydrogen and solar panels.
And the driving force behind these successes is massive government interventionism in the tech sector. Hence, the increasingly shifting focus of the CPC on advancing scientific and technological growth is a sign that China is looking to gain self-reliance amidst an external environment made uncertain by export bans from the US, the Netherlands, and Japan, disruptions in supply chains, and narratives on 'decoupling' and 'de-risking .'
In fact, at a politburo study session earlier this year, Xi said:
"Urgent efforts are required to accelerate the building of the country's self-reliance and strength in science and technology to solve the issue of the foreign "stranglehold."
And we can see this play out in the recent crackdown by Chinese authorities on foreign companies or companies with foreign clients, such as Capvision, Micron, and Bain and Co. To benefit its growth, China has either ceased its operations in the domestic market under the garb of 'counter-espionage' or has forced them to transfer their intellectual property tech to China. Either way, China is weaponizing access to its vast market base in exchange for political or technical benefits. It has also expanded the scope of how its laws define 'espionage' and 'state secrets' to prevent foreign companies and clients from stopping leaks of Chinese breakthroughs and initiatives.
The quest for self-reliance in critical and emerging technologies is misplaced, but the US-led West and China are targeting sectors to gain supremacy. Export controls and strategization of trade has led China to bolster roughshod scientific and technological growth, and its policy focus on the issue is only likely to expand as a new form of 'techno-nationalism' takes hold.
Cyberpolitik 2 : An IAEA type regulation for Artificial Intelligence?
— Saurabh Todi
Artificial Intelligence (AI) has recently found several applications, primarily through Large Language Models (LLMs) such as ChatGPT and Bard and AI-generative art platforms such as Midjourney. All these tools have found significant public traction in recent times, making these tools accessible to a large section of internet and social media users. However, the substantial potential in generative AI to produce deep fake images and spread misinformation has raised concerns about AI regulation. This concern gains prominence given the possibility of deep fakes to undermine electoral behaviour and flame ethnic tensions.
OpenAI, the organisation behind ChatGPT, recently proposed that a regulating agency similar to the International Atomic Energy Agency (IAEA) should be established to impose restrictions on AI deployment and ensure that AI tools and platforms adhere to safety standards. Interestingly, this proposal has already garnered an endorsement from the UN Secretary-General António Guterres. He also announced plans to form a high-level AI advisory body that regularly evaluates existing AI governance frameworks and provides recommendations.
It remains to be seen if this idea gets significant traction from the global community. However, any such mechanism must ensure adequate participation of all countries and civil society organisations during these discussions. Furthermore, any such instrument should go through the deliberative process of the UN General Assembly to allow for the broadest participation and to avoid the dominance of influential players in the global AI industry, including economically and geopolitically powerful countries.
Antariksh Matters : Artemis Foiled
— Aditya Ramanathan
NASA’s Artemis 3 mission - the one that is to put two humans on the lunar surface - is likely to be delayed until at least 2026.
This is hardly a surprising development. Artemis 3 depends on SpaceX’s Starship, a massive rocket-and-spacecraft combined that is supposed to be able to haul up to 250 metric tonnes on a one-way trip.
However, Starship’s maiden test flight on April 20 ended in a fireball after the rocket failed to separate from its payload. But that was not the only problem. Starship’s 33 first-stage Raptor engines pounded the launch pad, ripping off chunks of concrete and sending debris flying hundreds of metres. The clouds of dust and sand that spread around the Boca Chica launch site were also dramatic enough to prompt a lawsuit against the US Federal Aviation Administration for licensing the launch despite the risks.
SpaceX considers the launch a partial success since the massive rocket clearly did much of its intended job. However, given the environmental damage, it’s unclear when SpaceX will receive clearance for its next Starship test. These concerns prompted a NASA associate administrator, Jim Free, to indicate that the Artemis 3 launch date would likely be pushed to sometime in 2026 from December 2025.
NASA’s Artemis 3 mission involves two different types of large rockets locked in a cosmic dance to the lunar surface. One of these rockets is NASA’s own Space Launch System (SLS). The other is Starship. In the mission's first leg, a Starship with the lunar Human Landing System (HLS) will be put into a special ‘Near Rectilinear Halo Orbit’ around the Moon. This is the same sort of elliptical orbit intended to be used by the future Lunar Gateway project - essentially a small space station orbiting the Moon.
Once the HLS is in orbit, NASA’s SLS will launch four astronauts aboard the Orion spacecraft. The Orion will dock with the HLS, transferring two astronauts, who will land on the lunar South Pole, while the remaining two continue to orbit the Moon on Orion. After six lunar days of experiments and sample collection, the two astronauts will return to the Orion and return to Earth.
It’s an incredibly complex and ambitious mission, and delays are highly probable. We should not be surprised if even 2026 proves too ambitious for Artemis 3.
Cyberpolitik Explainer : Electronics Repair Services Outsourcing Pilot
— Rijesh Panicker
The Ministry of Electronics and Information Technology (MeitY) announced the start of a three-month pilot under the Electronics Repair Outsourcing (ERSO) initiative to validate process and policy changes made to incentivize the industry and make India the repair capital of the world. The pilot, being conducted in Bangalore, will see 5 companies, including global repair services firms Flex and CTDI, Lenovo, R-Logic & Aforeserve, participate. As part of this pilot, firms will be allowed to repair electronic devices bought in from other regions, with the expectation that they are re-exported. The pilot will validate streamlined customs processes for approval of import and export of electronic products (down to 1 day from 15 days) and relaxed e-waste rules which prevent the disposal of non-repairable e-waste domestically (during the pilot, 5% by weight of imports can be disposed of locally).
Globally, the rise of the repair outsourcing industry can be linked to the “Right to Repair” movement. This legislative movement seeks to make it easier for consumers to repair their products by mandating that manufacturers provide spare parts, provide repair information, and provide diagnostic tools. Various such legislations have been passed around the world. In Europe, manufacturers must provide spare parts for up to 10 years. The FTC has been mandated to draft right-to-repair legislation in the US, and some states have already passed legislation. India is also in the process of legislating a right-to-repair framework.
India’s hardware association (MAIT) has estimated that the industry could bring in $20 billion in revenue and 5 million jobs over the next 5 years. Currently dominated by companies in China, Malaysia, and Indonesia, the ERSO industry is estimated to be worth $100 billion and grow to $200 billion by 2031. India already has an ecosystem of independent repair service providers to domestic consumers. The availability of trained manpower and lower cost of labour are expected to be key competitive advantages. In the long run, the success of ERSO companies could make India a key market for electronic components and have a flywheel effect in attracting electronics component manufacturing to India.
Our Reading Menu
[Op-Ed] Free Bus Travel Is Not A Silver Bullet, by Suman Joshi.
[Report] Top Talent, Elite Colleges, and Migration: Evidence from the Indian Institutes of Technology, by Prithwiraj Choudhury, Ina Ganguli & Patrick Gaulé.
[Newsletter Piece] The Coming Wave of AI, and How Nvidia Dominates, by Doug O’Laughlin.
[Discussion Document] Analysing India’s Position on the Information Technology Agreement, by Satya S. Sahu, and Miheer Karandikar.