I think, therefore AI am

This is a summary of the authors’ insights and what’s been discussed during the webinar “I think. Therefore AI am” that took place on April 27th and which is viewable at

Generative AI is gaining funding from private and corporate entities and may lead to both enhanced productivity and long-term GDP growth. However, it also brings a range of social and ethical implications.

Generative AI has been gaining traction since 2014, though its origins date back as far as the 1960s. Last year, the release of Open AI's GPT-3.5 catapulted the technology into the public spotlight, and it may now have over 100m users, becoming the fastest growing start-up in history. Based on the transformer architecture, GPT is designed to solve sequence transduction problems, such as in language translation and speech recognition. It can also be used to generate realistic and creative content that was once considered the exclusive domain of humans. The underlying technology, ‘large language models’, are now seem by many people in tech as being a ‘once in a decade’ generational platform shift, comparable to the web or the iPhone – some people see its significance as even greater.

The semiconductor sector is the most obvious beneficiary of the implementation of generative AI and GPT technologies. In May, Nvidia said that its revenue for Q2 would be 50% above previous estimates, as every cloud platform scrambles for more computing capacity. Estimates place the global semiconductor market at around US$ 600 billion, with approximately one-third of the market expected to experience the impact of these advanced technologies. This implies that the market for dynamic random-access memory (DRAM) and graphics processing units (GPUs) will be estimated to have a value of around US$ 130 billion and 40 billion respectively. What’s more, given it is difficult to predict the number of servers that will be necessary to sustain public and private models, major corporations may begin to develop their own versions.
 

The semiconductor market is expected to reaccelerate thanks to the rise of AI. The PC wave, the internet wave, the smartphone wave and the cloud wave have all contributed to an accelerated growth of 7% since 2015-16, with forecasts predicting the market will likely reach US$1 trillion in 2030.

However, while the industry is set to grow, geopolitical tensions and uncertainty pose potential setbacks. The US, for instance, has imposed restrictions on the sale of US technology since 2018, in part, to block China's Made in China 2025 plan, which aims to give China independence in chip manufacturing. These initial restrictions were stepped up under the Biden administration, with the inclusion of Chinese firms on the US entity list.

In 2023, the US also banned the sale of advanced chip-making equipment to China and persuaded the Netherlands and Japan to implement similar restrictions.

AI is reshaping the media industry by revolutionising the creation of content – from music to publishing to video production. With AI, existing rights and materials can be used to generate new works; for example, books can be rewritten and lyrics composed based on a given context. Start-ups are already capitalising on the potential of AI to create content, and this trend looks set to continue in the immediate future.

In some fields, such as special effects and production tools for cinema or content production for games, there is immediate scope for greater efficiency, while generative behaviours also offer scope for much more engaging interactions with so-called ‘non-player characters’ in games.

On the other hand, in some fields there is also likely to be an explosion in content creation, which we have already been witnessing since 2018. Luminate reports that almost 120,000 new audio tracks are now submitted to streaming providers every day. AI can only accelerate this. It will become much easier to create music (as seen previously with, for example, synthesisers or drum machines), but this kind of dilution (often simply spam) could make it difficult for fans to find music. It also strengthens existing calls to change the way rights holders are remunerated. Labels and artists could therefore push platforms to implement a system of differentiated payments according to artist performance, profile or perceived value.

However, in the longer term generative music poses some hard IP questions. If I ask ‘play these lyrics using Taylor Swift’s voice’, the copyright issues are well-understood, but if I ask ‘create some music in the style of top 10 hits of the last 5 years’, it’s much less clear what royalty payments should be – if any.

Generative AI has the potential to play an active role in fighting climate change. Evidence suggests that AI patents with climate applications are growing more sharply than climate change-related inventions. This creates the opportunity for generative AI to provide more reliable climate specialist data than solutions available today.

Investors are increasingly looking to new AI solutions based on generative AI to identify investment opportunities. In terms of climate, the energy consumption of generative AI models is a topical issue. Fortunately, most of the energy consumption and pollution occurs during the training period, not the usage period, meaning its use does not directly lead to higher emissions. That said, there is still a lack of scientific consensus on the energy consumption of generative AI due to limited testing in lab settings. The energy consumption of AI infrastructure, and data centres in general, depends a great deal on where they are physically located and hence on their power sources, while the energy consumption of the models is also changing rapidly.

In addition, the social and ethical implications of generative AI are complex and can be seen as threatening. To better understand the effects of this technology, it is important to consider not only the jobs that may be lost, but also the impact it could have on employees. Those less qualified in particular are often affected by developments in generative AI. The general pattern of AI, like all automation in the last two centuries since the industrial revolution, has been that some jobs are lost in the short term and more are created in the longer term, with net employment remaining the same, but with frictional pain in the transition.

Meanwhile, an integral part of the previous wave of AI was understanding the ways that AI systems and their training data can unintentionally reflect or amplify existing biases within society, or make undesirable outcomes easier, much like databases before them.

To counter this, companies can set up training plans to teach employees how to use generative AI, or indeed create new jobs based on the technology. This is one of the most concrete impacts of generative AI in terms of ESG. And, with regard to ethical implications, UNESCO has already begun looking into the matter to ensure that basic human rights are not violated. The legal framework for AI is also likely to change in the future, though it is still difficult to predict what this will look like.

Ultimately, generative AI is a rapidly expanding field with a wide range of potential applications. It has the potential to revolutionise countless industries and increase productivity, however, it also brings with it ethical, legal, and environmental challenges. Companies must therefore ensure they are aware of the implications – both the positive and negative – associated with advances in this space.

Read also Eric Benoist’s thematic research report “I think, therefore AI am”