Steady growth | January predictions

AI technology will continue to grow both deeper, through powerful foundational models, and wider, through more connections to the rest of the digital ecosystem (e.g. the internet of things, virtual reality, digital infrastructures). In parallel, smaller AI and open source models will gain momentum as a more transparent, adaptable and environmentally friendly approach to AI. 

In 2024, Google will try to challenge ChatGPT’s dominance. It remains to be seen whether the AI space will be bipolar, with OpenAI/Microsoft and Google, or multipolar, with more actors joining at the top. In addition to Apple, Amazon, and other companies, the most interesting challenge for duopoly comes from Meta, which has been supporting the open-source LLaMA model. 

Large Multimodal Models (LMMs) will gain in relevance following the shift from text towards integrating video, images, and sound in AI foundation models. 

After realising the strategic relevance of AI-codified knowledge, companies, communities and countries will intensify the development of bottom-up AI based on in-house solutions. AI models will be localised through the use of data pruning, adaptive computing, and models for retrieval augmented generation (RAG). Bottom-up AI will become a new AI dynamic in 2024. 

Small Language Models (SLMs) are already a thing, but cost-efficiency and sustainability considerations will accelerate this trend as AI has a big carbon footprint due to the use of a lot of energy for processing data and the development of foundational models. 

After surviving an attack from the ‘extinction risk’ camp in 2023, open-source AI will continue to develop fast in the new year. Open-source AI tends to be smaller but more adaptable and customisable. LLaMA is 10 times smaller than OpenAI Large Language Model (LLM). If LLaMA or other smaller and open-source models (BLOOM and OPT) can achieve similar results to GPT-4, it can be a major shift in the AI landscape with more inclusion for organisations and individuals with much fewer resources. 

More clarity | January predictions

There will be more clarity in discussions about governing AI and dealing with associated risks. Existing risks (e.g., jobs, misinformation, biases) will receive more attention than existential risks (human survival). Exclusion risks based on a few AI monopoly companies will become more relevant. AI governance will become more specific and concrete, addressing computational power, data and knowledge, algorithms, and AI applications.

AI risks will dominate the governance and regulatory debate. The main shift will be from a heavy focus on extinction and long-term risks towards existing risks that AI poses to society, from misinformation to education and the future of jobs. In addition, governments will focus more on exclusion risks triggered by the AI monopolies of the major tech companies. 

The end of 2023, with the adoption of the EU AI Act and the publishing of the Interim Report by the UN Secretary-General’s High-level Advisory Body on AI, signalled more clarity in dealing with AI risks. The previous heavy focus on extinction risk is balanced with existing risks and midterm risks, such as monopolies of AI platforms. 

Existing risks (short-term) from AI developments include loss of jobs, protection of data and intellectual property, loss of human agency, mass generation of fake texts, videos, and sounds, misuse of AI in education processes, and new cybersecurity threats. We are familiar with most of these risks, and while existing regulatory tools can often be used to address them, more concerted efforts are needed in this regard.

Exclusion risks (medium-term) could be triggered by the centralisation of AI knowledge in the hands of a few powerful players. Ultimately, their monopolies could create a risk of exclusion: citizens, communities, and countries worldwide will be limited in their ability to use and benefit from common knowledge. Such monopolies could lead to a dystopian future for humanity.  Legally speaking, risks of such AI monopolies can be reduced via antitrust regulation and protection of data and intellectual property associated with the knowledge used for the development of AI models. 

Extinction risks (long-term) are based on the possibility of AI evolving from servant to master, jeopardising humanity’s very survival. After very intensive doomsday media coverage throughout 2023, these threats haunt the collective psyche and dominate the global narrative with analogies to nuclear armageddon, pandemics, or climate cataclysms. 

The dominance of extinction risks in the media has influenced policy-making. For example, the Bletchley Declaration adopted during the London AI Safety Summit heavily focuses on extinction risks while mentioning existing AI risks in passing and making no reference to exclusion risks.

The AI governance debate ahead of us will require: (a) addressing all risks comprehensively, and (b) whenever it is required to prioritise them, that decisions be made in transparent and informed ways. 

Dealing with risks is nothing new for humanity, even if AI risks are new. In the environment and climate fields, there is a whole spectrum of regulatory tools and approaches, such as using precautionary principles, scenario building, and regulatory sandboxes. The key is that AI risks require transparent trade-offs and constant revisits based on technological developments and society’s responses.

Computation level

The main question is access to powerful hardware that processes the AI models. In the race for computational power, two key players—the USA and China—try to limit each others’ access to semiconductors that can be used in AI. The key actor is Nvidia, which manufactures graphical processing units (GPU) critical for running AI models. 

With the support of advanced economies, the USA has an advantage over China in semiconductors, which they try to preserve by limiting access to these technologies via sanctions and other restriction mechanisms.

Data and knowledge level

There are two sets of challenges: firstly, copyright holders worldwide are questioning in courts the use of their creative work in the development of AI; secondly, facing a limited data volume of internet content and in an attempt to respond to privacy and intellectual property concerns, companies are using synthetic data generated by AI. This recursive loop triggers ‘AI inbreeding’ and an inevitable degeneration of the quality and relevance of AI models. 

Copyright: In 2023, the question of protecting the data and intellectual property used by AI platforms started gaining momentum. The EU AI Act requires more transparency about data used for AI models. Writers, musicians, and the photography industry started court cases against OpenAI for use of their intellectual property for training AI models. The New York Times’ lawsuit against OpenAI for copyright infringement is one recent example. Other companies, however, are watching and learning valuable lessons: Apple decided to conclude a $50m multi-year deal with the Times to license material for its AI models. 

In 2024, new ideas will flourish. For example, some lawyers argue that, analogous with copyright, a ‘learnright’ should be established to govern how AI uses content for training. Various tagging and watermark systems are discussed for distinguishing human-made from AI-made artefacts.

There are also arguments against strict copyright protection and the need to allow data scrapping in order to facilitate progress of AI.

Synthetic data: Most existing models are based on internet content, considered as the ‘ground truth’ for AI development. This initial input into AI models is already depleted. In 2024, there will be two major problems with acquiring data. Facing limited available data, AI companies started using synthetic data generated by AI. According to Stanford University researchers there is a 68% increase in synthetic data generated by AI on Reddit. 

Algorithmic level

In 2024, discussion on transparency, evaluation, and explainability will need to operationalised as it will be required by the EU AI Act and other legal instruments in the field of AI. AI governance will also focus on the relevance of ‘weights’ in developing AI models: how to highlight the relevance of some input data and knowledge in generating AI responses.

Transparency: There is very little transparency when it comes to AI models; we know little to nothing about, for instance, the data fed into the model or the weights that are given to parameters. Transparency is a precondition for evaluation and, ultimately, accountability in the digital realm. The lack of transparency in AI follows a general trend of reduced transparency in the digital industry and social media platforms. Twitter, Facebook, and Google provide less and less access to their operations. 

Evaluation: The EU AI Act and other regulations require the evaluation of AI platforms. However, there are numerous limitations in conducting evaluation, from lack of data to lack of common evaluation methodologies. In the USA, bipartisan efforts in Congress, such as the ‘AI Research, Innovation, and Accountability Act,’ aim to increase transparency and accountability for high-risk AI applications.  

Uses level

The focus is on the practical use of AI via apps and other services. For example, regulations would focus on the implications that the outputs of systems such as ChatGPT have in terms of human rights, security, and consumer protection instead of regulating the algorithms that generate such outputs. Like with traditional digital systems, responsibility for obeying regulation will be placed on companies that develop and operate AI systems, such as OpenAI/Microsoft, Google, Meta, and Amazon.

For a long time, one of the pillars of digital governance has been to regulate the uses and outputs of digital systems instead of regulating how the internet functions technically (from standards to the operation of critical internet resources like internet protocol numbers or the domain name system). This approach is one of the main contributors to the fast growth of the internet. The current calls to shift regulations on the algorithm level (under the bonnet of technology) would be a major departure from the well-tested and established approach of technology regulation, with far-reaching consequences for technological progress. 

In 2024, semiconductors will remain the main digital battleground between China and the USA. So far, the USA-led restrictions on the export of chip-making technology have triggered rapid growth in this sector in China. Alibaba, Xiaomi, Baidu, and other Chinese companies invest heavily in the local semiconductor industry. The same tendencies can be observed worldwide as countries try to become self-reliant in the development and production of advanced chips. 

The previously integrated network of submarine cables will continue separating between China and the USA. We have already seen, for instance, that the USA banned landing rights for a few submarine cables with links to China. And China started investing in separate cable systems, such as a new cable connecting China with France via Singapore, Pakistan, and Egypt. 

Faced with these divisions, countries will focus on digital sustainability by developing alternative digital routes. They will also work more on preserving content locally via internet exchange points (IXPs).  

Outer space is another field of accelerated competition between the USA and China, as well as other public and private actors. In the area of internet connectivity alone, low-orbit satellite (mega)constellations by the likes of SpaceX’s Starlink and Amazon’s Project Kuiper are to face competition from China Satellite Network Group’s Gouwang project. 

National space agencies and private actors alike are planning or working on deploying (new) communications and navigation networks (e.g. around the Moon), the development of new space stations, and the exploration and exploitation of space resources (e.g. metals, minerals), giving rise to new governance issues. 

Against this backdrop, ensuring a fair and responsible allocation and use of spectrum resources and orbital positions, encouraging responsible behaviour in outer space from public and private actors, and ensuring the sustainability of space activities, will be some of the questions that UN bodies and processes—from the Committee on the Peaceful Uses of Outer Space (COPUOS) and the International Telecommunication Union (ITU), to the Summit of the Future—will be addressing in 2024.

In 2024, there will be a push for new organisations, commissions, and expert groups dealing with AI and digital governance. ‘Governance inflation’ will be fuelled by media hype and confusion about AI as a ‘wicked’ policy issue. AI will be used by international organisations to carve out new mandates and ensure their ‘raison d’être’ in the future.

The adoption of the UN Cybercrime Convention will mark the beginning of global digital governance in 2024. The Global Digital Compact (GDC) will be negotiated ahead of the Summit for the Future in September 2024. The main challenge will be to align the GDC with the World Summit on the Information Society (WSIS) and the 2030 Agenda for Sustainable Development. In the turbulent year of 2024, the UN Security Council and other bodies will have to deal with digital aspects of conflicts and humanitarian crises.

2024 will be marked by an interplay between change, which is the essence of technological development, and continuity, which characterises digital governance efforts. Change in search of ‘the next big thing’ is at the core of tech narratives. Governance has different dynamics for a reason. The foundation for the current global digital governance architecture was set back in 1998 and remains largely relevant today. Decisions to bring changes in this space should be made with calm and clarity, allowing for a clear separation between governance that works and governance that needs to be changed with a reason. 

At the UN, the year will start with the adoption of the Cybercrime Convention. Most of the year will be dominated by discussions and negotiation of the Global Digital Compact (GDC) and synchronisation with the wider UN agenda of the World Summit of Information Society (and the upcoming WSIS+20 review), the future of Internet Governance Forum, and the Agenda 2030 for Sustainable Development. 

In 2024, Brazil’s G20 Presidency will continue IBSA (India–Brazil–South Africa) momentum after India’s G20 Year. Brazil announced a focus on the following digital issues: digital inclusion, digital government, information integrity, and artificial intelligence. Brazil will also host a NETMundial+10 in the first half of the year with the aim to ‘update the global discussion on internet governance and the digital ecosystem’ and review the 2014 principles and roadmap for internet governance.

2004 was the year of ‘soul searching’ of diplomatcy for its role and purpose. Diplomacy has been overshadowed in recent years by military and confrontational logic. Unfortunately, this trend is likely to continue in 2024, with no end to current conflicts in sight and looming new conflicts.

Diplomacy will begin a ‘soul searching’ for its role and purpose. Technology will play an important role in shaping the future of this ancient profession by facilitating representation, negotiations, and the peaceful resolution of conflicts, all of which are core diplomatic functions.

AI will put diplomacy to the test through language, a key diplomatic tool. The automation of summarising, reporting, and drafting using Large Language Models (LLMs) will have a significant impact on the diplomatic profession.

The decades-long saga around online encryption will gain new momentum in 2024. In their push to get more access to encrypted communication, some governments are proposing client-side (on-device) scanning of communications for illegal content such as child pornography. Once such content is identified, an alert is sent to law enforcement authorities. After the message leaves the device, it will enjoy end-to-end encryption. 

Encryption serves as a vital tool for protecting personal and corporate data, yet it also poses significant challenges to criminal investigations. Traditionally, governments have been pushing for backdoor access to encrypted content—via online platforms—while tech companies and human rights actors have been resisting. In 2024, a more nuanced debate is gaining ground around the notion of client-side scanning: using algorithms to scan messages locally (on-device) for illegal content, and enabling the reporting of red flags to authorities. 

Proponents—including some governments—argue that on-device scanning would preserve the end-to-end encryption principle, but still support law enforcement in their fight against crime. Opponents—including some tech companies like Meta—argue that ‘client-side scanning could mark the end of private communication online’, as the technology could also serve as a tool for mass surveillance. 

Digital identities will gain relevance in 2024. The most interesting development will be related to the Digital Public Infrastructure (DPI) initiative, which provides a solution for managing identities online. DPI was endorsed by G20 during the New Delhi Summit and gained relevance in international debates in 2023. 

Having a digital identity is becoming critical for individual participation in the economic and social life of digitalised societies. Digital identity is used for access to government, economic, and other services. Proper technical and governance solutions for digital identity involve a delicate interplay between the protection of privacy, access to societal services, financial transactions, and other aspects of our lives that require proof of identity.

Identity management opens many controversies. For example, the proposed revision of the EU regulation on electronic identification, authentication and trust services (eIDAS regulation) has attracted criticism over certain provisions that would allegedly enable governments to intercept web traffic without effective recourse, limit the security measures that could be taken to protect web traffic connections, and otherwise deteriorate privacy safeguards. 

In 2024, there will be more discussion on the interoperability of online identities between countries, as national jurisdictions over identities collide with the trans-border nature of most digital services and activities. 

In 2024, according to the Economist, 4.2 billion people in 76 countries will hold some sort of national election. Digital platforms and tools will play an important role in campaigns and elections. Their relevance increases given the high importance of these elections for the future of democracy itself.

AI empowers the generation of content, including fake and manipulative content such as the following deepfake of Trump and Fauci posted by Ron DeSantis.

Link: https://www.youtube.com/watch?v=hLuUmNkS21A

Some platforms like TikTok, Discord, and Twitch are developing new tools to handle election disinformation, while others like X and Meta have rolled back their policies. However, experts have already expressed concerns about platforms not having sufficient resources for monitoring complexity of online content during elections. 

In Michigan, USA, there is impending legislation to regulate AI in political advertising highlighting the need for transparency and accountability in how AI is leveraged during electoral processes.

Detection of AI-generated content

The race to create and detect AI content will speed up in 2024. Those who use AI to generate text, video, and sound in the race have a significant advantage. The “AI detecting” camp attempts to catch up in the race using two major approaches: detection and watermarking.

Content moderation

Major tech companies, such as Alphabet, X, Meta, and TikTok, will play an increasingly prominent role in content policy and moderation. They are becoming de facto content regulators, determining what content is allowed and what is removed from their platforms. In 2024, these companies are expected to continue adapting their content moderation policies to address the growing concerns around misinformation, fake news, and violent content.

Content regulation

The implementation of the EU’s Digital Service Act (DSA) will gain momentum in 2024. Analogous to the ‘Brussels effect’ from data governance, DSA is likely to be emulated in other jurisdictions worldwide. 

The internet has become an unregulated space where violent ideologies flourish unchecked. The US and EU’s joint statement on cyber resilience reflects a concerted effort to address cyberterrorism and online radicalisation.

Because governments and tech companies lack adequate policies and technical tools for content governance, the arbitrary prohibition of certain content could cause significant social and political unrest. Tensions will spill over into the streets and squares from online spaces. Following the crisis of 2024, more stringent content governance policies will emerge.

Content fragmentation

The digital space is fragmenting with the development of smaller and segregated online communities of like-minded people. For example, when Trump was banned from Twitter he moved to Truth Social platform which gathers users with similar views. 

This trend of fragmentation means a further disintegration of social spaces and ‘online squares’ with far-reaching consequences for social cohesion and political systems.

Transparency

Social media companies are becoming less and less transparent about their activities and business models. Most of them stopped and reduced access to their data to independent researchers. For example, X (Twitter) stopped free access to the platform’s API. Meta is also restricting access to their services. The EU’s Digital Service Act provides regulatory solutions via provisions to allow researchers to monitor social network platforms. The lack of transparency on social media platforms will be particularly problematic in monitoring online aspects of the forthcoming elections in 2024.

In 2024, AI will trigger new aspects of inclusion in addition to the traditional issues of internet access. The main issue is the incorporation of knowledge from various cultural, regional, and civilisation traditions in the development of AI models. Current AI models are based on limited datasets, primarily Western ones. In the coming years, communities will aim to develop bottom-up AI solutions that reflect their cultural and knowledge heritage.

Inclusion is a cornerstone principle of the 2030 Agenda for Sustainable Development, and one that should guide all our efforts to ensure that no-one is left behind in the march into a brighter global future. We must make certain that all citizens, communities, and nations benefit from the historic transition to a digital world, and that special attention is paid to those groups that have historically been neglected or ill-served by technological progress, such as women and girls, those with disabilities, youth and children, and indigenous peoples.

In the 2020s, the challenges of digital inclusion will demand a holistic approach that is able to take into account all of the following policy areas: 

• Access inclusion: equal access to the internet, information/content, digital networks, services, and technologies.
• Financial inclusion: access to affordable and trustworthy digital financial and banking services, including e-commerce, e-banking, and insurance.
• Economic inclusion: facilitate all individuals’, groups’, and communities’ ability to participate fully in the labour market, entrepreneurship opportunities, and other business and commercial activities.  
• Work inclusion: support and promote equal access to careers in the tech industry and elsewhere irrespective of gender, culture, or nationality. 
• Gender inclusion: educate and empower women and girls in the digital and tech realms.
• Policy inclusion: encourage the participation of stakeholders in digital policy processes at the local, national, regional, and international levels. 
• Knowledge inclusion: contribute to knowledge diversity, innovation, and learning on the internet. 

As we endeavour to find unified responses in this varied range of spheres, and as we are forced to make informed trade-offs between different goals and interest groups, clarity in our thinking about statistics and policy will be essential. Without it, we will be negligent in our duty to work towards the digital inclusion of the ‘next’ or ‘bottom’ billion of digitally excluded citizens of the world.