AI – beyond the hype
Features
In Our Latest Future Now Issue we delve Into AI Innovations That Are Truly Making a Difference in Real-World Applications
It all seemed so world shaking. In November 2022, ChatGPT shocked the world with its versatility and eerily human-like responses to almost any question. Do you still remember the first time you used it and what you thought?
The rise
‘That’ ChatGPT moment fired the starting gun on an extraordinary period of focus on generative artificial intelligence – a category of technologies that includes the large language models that power Open AI’s flagship consumer product and its competitors. Since then, Mckinsey has estimated that generative AI could add the equivalent of $2.6 trillion to $4.4 trillion to the global economy – each year. And, in June 2024, Nvidia surpassed Microsoft as the world’s most valuable company as its graphics processing units, the key piece of hardware in AI-focused data centres, became the belle of the stock market ball. Hopes and fears for the technology have been so acute that the world’s dignitaries have rubbed shoulders with tech executives at high-level AI summits at the UK’s Bletchley Park and (virtually) in Seoul.
The fall?
Fast forward to today, however, and it has already become fashionable to see the whole thing as one big hype bubble – and with some cause. In July, The Economist highlighted that despite the AI mania, the technology has had ‘almost no economic impact.’ And, at the end of July, stock markets fell sharply as investors unloaded shares in AI-heavy tech firms, partly driven by uneaseabout the extraordinary levels of capital expenditure these businesses have committed to as they keep pace in the AI race.
The planetary impact of the accelerating expansion of power-thirsty data centres, which underpin AI, has also added to the souring mood, with Google and Microsoft reporting increases in emissions as a result of AI infrastructure.
The long term
Is this scepticism ultimately warranted? Betting against AI over the long term would be foolish, especially as the story we are seeing today has played out before. In 1987, economist Robert Solow quipped that that the computer age was everywhere except for the productivity statistics, foreshadowing, almost exactly, some of the arguments of today’s AI bears. And in May 2024, when the tide of sentiment had already begun to turn, Goldman Sachs research found that AI was showing very positive signs of *eventually* boosting productivity and GDP.
For most people generative AI has been synonymous with chatbots – but there is more to the technology. To cut through the Gordian knot of AI boosterism on the one hand and nay-saying on the other, we’ve therefore focused this month’s Future Now on innovations in generative AI that get behind the headlines to address less-discussed applications for the technology.
Beyond human language
While most of the discussion around generative AI has revolved around chatbots that generate responses to prompts in human language, this is not the only way that the technology can be used. Indeed, generative AI is showing promise in materials science and drug discovery, where it is molecules or protein sequences, rather than words, that are being generated. McKinsey, for example, forecasts that the technology could generate $60 billion to $110 billion in economic value for the pharmaceutical and medical product sector each year.
The innovations:
Photo credit: Cradle Bio
AI optimises protein sequences for enhanced products
Proteins are crucial building blocks in biology, and increasingly show promise for creating new products – from medicines to new foods, detergents, and petrochemical alternatives. Proteins are made up of sequences of amino acids, with the exact sequence determining the protein’s properties. And, just as generative AI can be used to string together chains of words, so it can be used to generate these protein sequences. This is what startup Cradle Bio is doing with its easy-to-use platform. The proprietary generative AI enables users to create variants of target protein sequences with improved characteristics, such as greater thermostability or optimised gene expression. This can all be done by a non-specialist at the touch of a button. Find out more
Photo credit: © ภาพของThanapipat Kulmuangdoan via Canva.com
AI-enabled protein design for gene-editing
Profluent is another startup that is using generative AI to de-code the ‘language of biology.’ Traditionally, researchers designing proteins for pharmaceutical or other use cases, either had to search for existing proteins in nature, or make small tweaks to natural proteins in the hope that they will deliver the desired properties. However, Profluent’s protein language models – which are similar in principle to the large language models used in chatbots – can be used to precisely design proteins and optimise their properties. The company is initially applying its platform to gene-editing, where genetic material of a living organism is altered by adding, deleting, or replacing sequences of DNA – often to cure a genetic illness. In April, the company announced that it had successfully edited the human genome using a gene-editor developed using its AI models. Find out more
Photo credit: © Julia Filirovska from Pexels via Canva.com
A generative AI search engine for advanced materials
The world will a require a range of new technologies and materials to tackle climate change. However, traditional testing processes can be time-consuming, expensive, and wasteful when things don’t go to plan. Now, CuspAI is applying generative AI to he problem, with its search engine that allows users to request and evaluate the properties of existing and new materials on demand – users could request a material that selectively binds carbon dioxide under specified conditions for example. The platform then uses generative AI, deep learning, and molecular simulation to generate, evaluate, and optimise potential molecular structures that meet those exact criteria. Read more
Streamlining scientific research
According to the Royal Society, “the applications of AI in scientific research are bringing a new age of possibilities and challenges.” And while there are obvious potential pitfalls for generative AI use in research, such as the risk of hallucinations and bias and the possibility of fraud, the technology can help humans to plan, conduct, and report on research in a better way. Innovators are therefore developing specialist tools that help with every stage of the process, from understanding the existing literature to writing up results.
The innovations:
Photo credit: Elicit
AI finds relevant research for time-poor academics
Reviewing past research is essential for academics, but it can be a slow process, as time is often wasted trying to source appropriate articles or reading content that ends up being irrelevant. Elicit is an online resource that utilises state-of-the-art AI to streamline academic research, acting as a virtual research assistant. Users type their research questions in plain English into the Elicit search engine and the platform draws from over 125 million articles in its database to create a personalised list of relevant papers, including those involving related keywords. Helpful one-sentence abstract summaries allow readers to quickly identify the most relevant resources and findings. Elicit claims that its tool can help users save five hours every week, cutting in half the amount of time it would normally take to do the same amount of research. Read more
Photo credit: © Mathias Reding from Pexels via Canva.com
Using AI to bring academic research to the fore
AI can help prioritise publications based on an individual’s preferences, which is how social media platform feeds work. And for those worried about increasingly narrow perspectives forcing people into echo chambers, a new discussion platform is bringing the latest research to the forefront of public discourse. Created by a Danish startup, the AI-powered platform, called Proemial, personalises reading recommendations to users and suggests connections across many different fields of study. The company brings together a variety of AI models to digest research and then make it not only applicable to scholars, but also interesting, relevant, and easy to understand for the general populace without specialist knowledge. Read more
Photo credit: © Ake Puttisarn from Getty Images via Canva.com
‘AI scientists’ that run their own research
In Japan, Sakana AI has recently introduced its ‘AI Scientist’, which allows large language models (LLMs) to independently perform machine learning research and speed up the entire process. The AI Scientist automates the whole research lifecycle – it can generate research ideas, execute experiments, summarise experimental results, and present findings in a full scientific manuscript. The system can even automate the peer review process to review the papers generated, write feedback, and develop ways to further improve results. While the AI Scientist demonstrates a strong ability to innovate on top of well-established ideas, such as diffusion modelling, it’s still yet to be seen whether it can also come up with novel, paradigm-shifting ideas. Read more
Giving healthcare a boost
Another area where generative AI can help to ease pressure is in the healthcare industry, where it can be used to streamline back-office processes and make time-poor clinicians more productive. In a 2024 Mckinsey survey that included 100 US healthcare leaders, 73 per cent of respondents believed that generative AI could be a boost for clinician productivity, with 60 per cent reporting that it could help with administrative efficiency and effectiveness. There have been several attention-grabbing headlines about the potential of general-purpose AI models in the healthcare space. For example, in February 2023, it was widely reported that ChatGPT had passed the US medical license exam. Researchers from Uppsala University have also found that one in five UK doctors are already using AI chatbots in their clinical workflows. However, innovators are working on more specialised tools that meet the particular needs of the healthcare industry.
Photo credit: © Alena Shekhovtsova from Baseimage via Canva.com
Speech-to-text tech for doctors in Africa
Healthcare systems across Africa are often particularly overstretched, with doctors in these countries frequently seeing far more patients each day than is typical for clinicians in other areas of the world. Voice-to-text apps can help to ease the pressure, but existing technologies often aren’t designed with African dialects and accents in mind. Now, Nigerian health technology company Intron Health has built a solution. The Transcribe app helps doctors complete clinical notes up to seven times faster than if they were typing, and has been built to accurately translate real-time speech and healthcare terms for more than 200 African accents. So far, the app works at a 92 per cent rate of accuracy for the included accents. Read more
Photo credit: © sasirin pamai’s Images via Canva.com
AI provides doctors with instant answers to clinical questions
To provide the best possible care for patients, doctors need to stay up to date with the latest clinical evidence and research as they make their diagnoses. Now, Medwise has developed an AI-powered search engine designed specifically for use by medical professionals. The sources for the platform include national knowledge resources as well as patient information and local guidance, all of which is vetted by clinicians. Users of the platform can then ask questions and get bite-sized answers – not full documents – at the moment they are caring for patients. Find out more
Photo credit: © Negative Space from Pexels via Canva.com
A decision support AI tool based on reputable medical evidence
Established businesses, as well as startups, are tackling the need for AI-generated answers to be drawn from robust and credible sources. For example, Elsevier Health, a global leader in medical information and analytics, is adding the convenient Q&A capabilities of generative AI to ClinicalKey, its clinical reference and decision support tool that is already embedded in thousands of hospital systems. Using the new tool, physicians can access Elsevier’s underlying body of curated and continuously updated medical content in a chat format. Doing so, they can be confident that responses are based on trustworthy medical information, with citations and underlying documents listed. Find out more
As a member you have unlimited access to our library over 14,000 innovation case studies. If this feature has piqued your interest, why not have a deeper dive into AI today?
23rd September 2024