AI Summit London 2024: highlights

Business leaders gathered at the AI Summit to discuss how artificial intelligence (AI) is accelerating productivity and innovation and even reshaping creativity. Technologists, executives and investors gathered in London’s Tobacco Dock in June for the eighth iteration of the event. Launched in 2016, it has grown in scale and scope, and now encompasses seven shows worldwide. Just over 4,000 people from 76 countries attended two days of this latest edition, more than a quarter of them business leaders.

Below we outline the latest on AI transformation, the evolution of creativity, what’s coming next in generative AI and wrap up with key takeaways from the show.

AI transformation

Across two days of talks, panels and presentations, leaders outlined the ways in which they are adopting AI and how their deployments are delivering gains. A seeming Swiss-army knife for businesses, here are five ways it is currently adding value.

Productivity: AI tools are now assuming or supporting on tasks in a broad range of contexts, from analyzing legal contracts to automating drive-through ordering systems. At Reuters it is helping free up time for “old-fashioned “shoe-leather” journalism by speeding up formerly laborious jobs like video editing and analysis. Catriona Campbell, UK&I chief technology & innovation officer at EY, said that the consulting giant’s £1.4bn investment in AI is “changing the very nature of the way we deliver our work” and that its 60,000 software engineers alone had seen productivity savings of 40%.

Optimization: Businesses are also using AI to optimize everything from resources to supplies. Centrica is using multiple data streams and large language models (LLMs) to inform investment and site selection for renewables infrastructure. Unilever is scaling the use of AI and machine learning to optimize product formulations, identifying ingredients to replace those which are expensive, scarce or unsustainable.

Accelerating R&D: Unilever is also using a range of techniques including custom LLMs, molecular simulation environments and generative chemistry to generate new chemistry in its research and development. Applications include identifying new molecules, assessing whether new products are microbiome-friendly or ensuring product safety without resorting to animal testing.

Better customer service: Speakers shared how AI chatbots are taking on the work of customer service agents at scale. Klarna’s bot, which handles service requests including refunds and returns, is reportedly doing the work of 700 full-time agents, delivering $40m in savings. But chatbots are going beyond efficiency to uplevel engagement. Natwest’s banking chatbot Cora, with 55m interactions under its belt, is getting a genAI upgrade and will soon offer personalized and proactive assistance, anticipating customer needs based on what it knows from past interactions. Brands are developing shopping assistants to offer help, guidance and advice, including Mastercard’s Shopping Muse and L’Oréal’s Beauty Genius (see below), which can diagnose, recommend and educate users who need more advice and guidance.

Faster content creation: Brands are now building in-house capability for swifter content creation. Jean-Paul Paoli, GenAI business transformation director at L’Oréal unveiled ‘Creaitect,’ an in-house GenAI beauty content lab which is being used to experiment in content production. The Creaitect suite of tools are trained on L’Oréal’s beauty brand visual codes which allows them to generate and localize high-quality compliant imagery for marketing. Digital commerce executives from Mars Wrigley shared that generative AI tools are allowing their marketing teams to compress production timescales for lifestyle visuals down to a week from several months.

BEAUTY GENIUS PRESS KIT IMAGE Credit L Oréal LD — Beauty Genius. Courtesy of L'Oréal.

The curator economy

Creativity is undoubtedly being reshaped by AI, but the debate rumbles on as to whether machines themselves can be creative. At two lunchtime sessions focused on AI in music, the panels predicted that the growing creative influence of AI will usher in a curator economy, where humans become tastemakers, rather than the authors.

Musical artist, writer and producer Daniel Bedingfield demonstrated how easy is to compose music today, spinning up a drum and bass tune in minutes using the generative AI text to music model Udio. The latter was developed by Google DeepMind and is currently the subject of a copyright infringement suit alongside a similar platform called Suno.

MIRO Daniel Bedingfield Fernando Garibay Ali Hossaini

Daniel Bedingfield The AI Summit London 2024 1475

1. MIRO, Daniel Bedingfield, Ali Hossaini and Fernando Garibay. 2. Daniel Bedingfield. Images: The AI Summit London 2024.

His co-panellist, the LA-based music producer Fernando Garibay, argued that we should redefine our concept of authorship in the age of AI-generated content. “We have to decouple our identity from our output” he said, “that will no longer serve us.” Instead, as curators, the panel suggested, our role will be to guide and direct the output as well as to experience it. Ali Hossaini, co-director of National Gallery X, which explores the intersection of art and technology, offered that the essence of art lies in our encounter with it, not just the final product itself.

Either way, the panel concurred with much of the discourse already circulating in this space, concluding that humans and technology will be collaborators, and each will play to their strengths in the creative process. Perhaps AI, in its ability to surprise and inspire us, can even unlock entirely new forms of creativity?

A brief but illuminating presentation from filmmaker Simon Ball of film and video production company Blake House illustrated this potential for novel art forms. Ball shared clips from an award-winning short called Original Short Film which he directed alongside his creative partner Ieva Ball. Inspired by the hallucinations of AI, the pair explored a new practice in which they used batch processing within Stable Diffusion to generate images that sampled different artists, creating a series of dreamlike visuals.

Colin Jarvis The AI Summit London 2024 372 — Colin Jarvis. Courtesy: The AI Summit London 2024.

Future generative revolution

What’s next in the fast-evolving AI space? The appropriately named Colin Jarvis of OpenAI (job title: Distinguished architect), outlined four ways we can expect to see generative AI progress.

The ongoing arms race between model developers will spur advances in textual intelligence – the way computers understand and generate language. Jarvis shared data which shows that there is little to distinguish the capabilities between the top ten models, so the quest is on for the next big leap in progress.

As well as getting smarter, models will also get cheaper and faster, said Jarvis. AI is following the typical tech trajectory of increasing power and declining cost, and intelligence per dollar is falling. For businesses this means that use cases that were previously out of scope due to cost may soon become viable.

We are also likely to see more model customization, as companies create their own bespoke, specialized models by applying sophisticated post-training techniques to base models.

As noted in VML Intelligence’s coverage of SXSW 2024, the next phase of generative AI will be multimodal, not just text based. Instead of having to stitch models together, we will see single models that are capable of both ingesting and outputting data in multiple modalities, including code, audio, image and video. As an example of the kinds of applications this might unlock, Jarvis showed a video clip of an AI-powered app called Be My Eyes which uses image processing and voice technology to help blind users understand their surroundings.

Six takeaways for brands

Like digital transformation, AI adoption and implementation is challenging and requires much more than the technology itself to be successful. Below are some of the learnings and recommendations VML Intelligence picked up across this year’s Summit.

Build for AI’s future capabilities: Don’t build for what AI can do now, but what it will be able to do in the future, says OpenAI’s Colin Jarvis. Development is changing so fast that by the time you have built something, the field will have moved on.

AI alone will not differentiate: AI is rapidly becoming table stakes and the technology itself will not differentiate. The user experience, proprietary data and services you bring to it will deliver difference, said Jarvis.

Balance human control and automation in user experiences: Speakers pointed to the need for AI to hand off to human expertise in certain cases, where processes or decision-making are more complex or where more empathy is required. Tim Bond, associate director, Media at IPSOS quipped that AI is an “exoskeleton that needs a human at the heart, otherwise it is just a bag of bones.”

Don’t chase novelty, have clear objectives: Sid Chaudhary and Sawat Choudhury of Mars Wrigley’s digital commerce practice suggested that marketers can often be guilty of “chasing the shiny toy” rather than prioritising outcomes. Fractional CTO Adil Asif said that many AI adoption projects fail due to a lack of clear objectives or flawed decision-making. It’s important to remember that AI is not software, not iterative or sequential he said, “it’s closer to discovery than development and that’s how we as business leaders should conceptualise it.”

Consider unexpected impacts: Daniel Hulme, chief AI officer at WPP, cautioned that AI won’t solve all your problems. “Just because you can doesn’t mean you should,” agreed Adam & Eve’s Sara Chapman. Both suggested it’s important to consider the knock-on effects of AI – either on the supply chain, or on teams. As Chapman points out, “simple tasks are what we normally give to juniors,” while also offering respite from high-octane work. Do we want to automate everything that is boring?

Make it ethical, transparent and accountable: As businesses set up task forces, draft key principles and assemble boards to oversee AI adoption, responsible AI is going mainstream, but there is more to do. A key focus should be on transparency said Daniel Hulme who noted that “AI is not currently explainable,” and that it needs to be easier to discern intent. Emma di Orio, global data privacy head at Diageo, argued for data privacy to be viewed as a business enabler, as a route to trust and a competitive advantage. Final word to Alyssa Lefaivre Škopac from the Responsible AI Institute, who agreed that “trustworthy Ai is going to win in the market.”