June 2023
A guide
to generative AI
Sponsored by
Produced in partnership with City Road Communications
June 2023
A guide
to generative AI
Sponsored by
Produced in partnership with City Road Communications
What is generative AI?
Generative AI has exploded into the mainstream. It is changing the way people work and disrupting entire industries. It also comes with challenges and risks that the business world will need to grapple with. An important first step for businesses is understanding exactly what generative AI is, and how it works.
Generative AI models are trained to identify patterns in enormous datasets to predict the output most likely to match a prompt.
These complex neural networks are known as foundation models, and can be adapted to perform specific tasks.
Large language models, or LLMs, are a subset of foundation models that focus specifically on natural language processing, with ChatGPT the best-known example.
At its most basic level, generative AIs are statistical models. A very basic example is a model that predicts the next word in the sequence ‘the cat sat on the’ is ‘mat’. However, the LLMs take this to a gargantuan scale – such as the entire contents of the internet.
The fundamental technology underpinning today’s generative AI tools has been around for some time. But rapid advances in training and the availability of computing power have turbocharged the AI sub-sector.
The result is systems that can provide human-like responses to any question, or generate realistic art, in just a few seconds.
Generative AI models can use various types of deep learning, including generative adversarial networks, diffusion models or variational autoencoders.
These systems are trained using two techniques: unsupervised learning and supervised learning.
Supervised learning requires labelled data, which tells the model what that information is. Over rounds of training, the model recognises patterns so that it can identify the data without annotations. This involves more human oversight at the earlier stages and while it is more labour intensive, it allows more manual fine-tuning of the results.
Unsupervised learning uses unlabelled data. The AI network is left to identify patterns and relationships on its own. This involves the AI model attempting to mimic the data and using the error in this mimicked data to correct itself, based on probabilities.
While this approach requires less time and labour, it can lead to the model making inaccurate predictions and requires larger training datasets.
Steve Harris
Head of Technology, Business & Commercial Banking, Lloyds Bank
“Putting sound business and financial plans in place alongside this new technology will be crucial to support long term success.”
"What the Midlands has done well is capitalising on the industries the regional economy was built on," says Aaron Baker, investor at BGF. "The automotive sector and its entire supply chain has seen exceptional tech innovation in the region – from batteries to electric vehicles – and that's because of the talent, experience, and rich heritage here. And this has evolved into global leading verticals in the sector, such as Leamington Spa's dominance in the gaming industry."
Generative AI in numbers
Projected value of the global generative AI market by 2032 | Source: Precedence Research
Monthly active users of ChatGPT in January 2023 | Source: UBS/Similarweb
Amount of funding raised globally by generative AI startups in Q1 2023 | Source: Pitchbook
Increase in demand for generative AI expert roles in the UK by job listings since launch of ChatGPT | Source: Adzuna
300 million Number of jobs that could be replaced by generative AI | Source: Goldman Sachs
Types of generative AI
Audio
+
Synthetic data
+
Key UK generative AI companies
Turns text prompts into video presentations given by AI-generated speakers that mimic humans | Founded: 2017 | HQ: London | Funding: $66.6m
Developer of AI models that create images, video and audio based on user descriptions | Founded: 2020 | HQ: London | Funding: $110.5m
Builds tools for businesses to create realistic human-like voice assistants | Founded: 2017 | HQ: London | Funding: $68.4m
Generates original musical compositions based on any genre | Founded: 2016 | HQ: Salisbury | Funding: $14.5m
A tool for the legal sector to quickly generate and review contracts | Founded: 2019 | HQ: London | Funding: $16.9m
PapercupCreates translations to existing content in a variety of languages with realistic spoken voice audio | Founded: 2017 | HQ: London | Funding: $33.2m
Why generative AI hype is (mostly) justified
As is the case with any trending technology, generative AI has faced questions over how much of the hype is justified. Critics have seen hype in areas like the metaverse and web3 come and go, making it fair to question whether generative AI will fare differently.
The metaverse, for example, generated a brief buzz thanks to Facebook’s rebrand to Meta. But when it came to delivering products that solve actual problems, very little has been offered and the excitement has since died down.
Something like the metaverse is based around speculation and vague concepts that do not meet an immediate business or consumer need. However, the latest wave of generative AI tools has come straight out of the gate with user-ready products that are already making an impact in the world of business and people’s day-to-day lives.
Ian West
Partner, Head of Tech Alliances and Tech, Media and Telco Sector, KPMG in the UK
“This latest form of AI is also much more powerful than that previously available and trained on LLMs with trillions of parameters.”
"What the Midlands has done well is capitalising on the industries the regional economy was built on," says Aaron Baker, investor at BGF. "The automotive sector and its entire supply chain has seen exceptional tech innovation in the region – from batteries to electric vehicles – and that's because of the talent, experience, and rich heritage here. And this has evolved into global leading verticals in the sector, such as Leamington Spa's dominance in the gaming industry."
AI as a general concept has been around for decades, and automation has been used in business for years. The difference with the dawn of generative AI is that it is highly accessible, easy to use and is improving rapidly with the addition of new data and training.
It also helps that the viral popularity of OpenAI products, including DALL-E and ChatGPT, has brought the capabilities of the technology to a much more mainstream audience than previous iterations of AI were able to do.
There is a long way to go for the technology to reach its full potential, and the pace of change is rapid. While the opportunities are vast, it is still important to keep in mind the challenges and limitations of generative AI. This will ensure businesses don’t get caught up in unrealistic expectations.
Steve Harris
Head of Technology, Business & Commercial Banking, Lloyds Bank
“As a result, we are seeing more exciting UK businesses take advantage of this technology. We continue to support our clients’ contributions to the large ecosystem of apps developing around these services, through a range of solutions and tools.”
"What the Midlands has done well is capitalising on the industries the regional economy was built on," says Aaron Baker, investor at BGF. "The automotive sector and its entire supply chain has seen exceptional tech innovation in the region – from batteries to electric vehicles – and that's because of the talent, experience, and rich heritage here. And this has evolved into global leading verticals in the sector, such as Leamington Spa's dominance in the gaming industry."
How generative AI will transform work
The creative sector is considered one of the most at risk to the disruptive force of generative AI. In just a few months, writers, photographers, designers and filmmakers have seen artificial intelligence create content that might normally take hours or days in a matter of seconds.
Big-name brands such as Heinz and Nestlé have already started using generative AI content in advertising campaigns. For companies, this could drastically reduce costs. Yet for workers who have spent years honing their skills, it poses a threat. Research from Goldman Sachs suggests that generative AI has the potential to automate 26% of work tasks in this sector.
However, it is unlikely to be a zero-sum game. Generative AI can also support creatives by automating laborious parts of their jobs, leaving them to focus on the parts that only a human can do. Similarly, AI could play a supporting role in the creative industry. For example, a marketeer could prompt Chat-GPT to suggest ideas for campaigns, then select and refine the best suggestions.
Generative AI has the potential to transform both the front and back ends of the financial sector. It could turbocharge personalised banking services, power conversational banking apps and augment internal banking functions.
Financial institutions have been using one form of generative AI long before the recent hype wave: synthetic data. This artificially created data lets banks study fraud and other negative financial events without the challenges associated with acquiring and handling personal data.
Despite the potential efficiency boosts, risk-averse financial service firms, particularly banks, are taking a more cautious approach to integrating generative AI tools. The Bank of America, Deutsche Bank and JPMorgan Chase are just some of the financial service firms that have restricted employees’ use of ChatGPT over concerns about the way it handles sensitive information. Financial service firms will also need to grapple with the explainability of customer-facing generative AI-powered tools, such as chatbots giving decisions on loan applications.
Sam Taylor
Director – Strategy & Performance Transformation, KPMG in the UK
“While AI can provide analysis and generate plans, the practical implementation, stakeholder engagement, and cultural adaptation are areas where human consultants shine. Generative AI should be seen as a force-multiplier in the sector, enabling faster and more value-added service delivery.”
"What the Midlands has done well is capitalising on the industries the regional economy was built on," says Aaron Baker, investor at BGF. "The automotive sector and its entire supply chain has seen exceptional tech innovation in the region – from batteries to electric vehicles – and that's because of the talent, experience, and rich heritage here. And this has evolved into global leading verticals in the sector, such as Leamington Spa's dominance in the gaming industry."
Generative AI has the potential to improve legal workflows. Companies like Robin AI have created tools to automate the process of drafting legal documents, such as wills and contracts. In this scenario, a legal clerk can check the auto-generated document and free up resources of often time-poor legal experts in the process. There is also scope to integrate generative AI into discovery, with legal professionals consulting a ChatGPT-style interface to ask if the file meets the criteria for a specific case.
More than 80% of law firm workers said they believe generative AI can be applied to legal work now, according to a Thomson Reuters Institute report. Despite this, generative AI in the legal sector is currently plagued by ethical dilemmas. Similar to banking, combining the legal sector with AI poses huge potential privacy issues.
There is also the very real risk that generative AI’s inaccuracies harm the legal process. One New York lawyer consulted ChatGPT to supplement legal research in a personal injury lawsuit. The judge found that six of the cases cited “appear to be bogus judicial decisions with bogus quotes and bogus internal citations”.
Alastair Mitton
Partner, Womble Bond Dickinson
“In parallel, it can break down legal jargon, explain judicial procedure and help with completing a large range of standard form legal documentation. On the other side of the fence, it is inevitable that it will form a key plank in creating efficiencies in the ways lawyers currently work.”
"What the Midlands has done well is capitalising on the industries the regional economy was built on," says Aaron Baker, investor at BGF. "The automotive sector and its entire supply chain has seen exceptional tech innovation in the region – from batteries to electric vehicles – and that's because of the talent, experience, and rich heritage here. And this has evolved into global leading verticals in the sector, such as Leamington Spa's dominance in the gaming industry."
Healthcare has a wide range of uses for generative AI, from automating time-consuming administration tasks, to suggesting testing options and discovering potential new treatments. At the macro level, generative AI can create models of large-scale health scenarios, such as the evolution of viruses to help plan for the next pandemic. AI can also provide educational assistance to train medical staff.
Like with other sectors, healthcare businesses will need to be wary of bias and inaccuracies in generative AI programs. Generative AI-powered chatbots can provide therapy without a human practitioner, which could improve patient access – after all, an AI therapist can be available 24/7.
These systems could scan a patient’s records and respond in a human-like way with advice. However, companies incorporating these tools will need to take extreme caution to ensure these chatbots are free from errors that could lead to real-world harm in a healthcare scenario.
While still not perfect, programmers are already using AI to almost instantly write up simple lines of code that would otherwise take real time away from more important tasks. Generative AI can also be used for testing, error checking, and as a source of suggestions for problem-solving.
Automation in software development, particularly low or no-code solutions, is not a new phenomenon. But the accessibility of the latest generative AI tools drastically improves its reach. For businesses struggling with the long-running shortage of programmers, generative AI tools can help them do more with less. Tools like GitHub Copilot can help coders brainstorm new ideas to solve coding challenges, shifting them from rote writing to strategic decision-making.
ChatGPT can write code in 11 different programming languages and, as the technology improves, coders will be able to lean on generative AI for more complex and time-consuming work, potentially even giving way for people without extensive training to get involved in coding.
Challenges and risks
Steve Harris
Head of Technology, Business & Commercial Banking, Lloyds Bank
“It’s important that businesses understand and manage these risks, as well as other risks which may come with running a fast-growth AI business, including managing large investment deposits and, where appropriate, foreign exchange risk.”
"What the Midlands has done well is capitalising on the industries the regional economy was built on," says Aaron Baker, investor at BGF. "The automotive sector and its entire supply chain has seen exceptional tech innovation in the region – from batteries to electric vehicles – and that's because of the talent, experience, and rich heritage here. And this has evolved into global leading verticals in the sector, such as Leamington Spa's dominance in the gaming industry."
Jobs
+
Technological
+
Copyright
+
Rose Smalley-Gordon
Managing Associate, Womble Bond Dickinson
“The ease with which many current generative AI tools can be used creates a whole range of potential issues. Just from an IP perspective, these include significant legal considerations around whether the content used to train the large language models was used in a non-infringing way.
“These concerns have also delayed the implementation of new regulations in the UK given the breadth of the infringement exemptions proposed to cover precisely this point. We have also seen litigation through to the Supreme Court (as well as top courts in other jurisdictions) which demonstrates current legislative provisions on IP ownership aren’t likely to be sufficient in this area.”
"What the Midlands has done well is capitalising on the industries the regional economy was built on," says Aaron Baker, investor at BGF. "The automotive sector and its entire supply chain has seen exceptional tech innovation in the region – from batteries to electric vehicles – and that's because of the talent, experience, and rich heritage here. And this has evolved into global leading verticals in the sector, such as Leamington Spa's dominance in the gaming industry."
Deep fakes
+
Leanne Allen
Partner, Data Science & AI Capability Lead, KPMG in the UK
“There are further risks that businesses must consider, one of which is ethical AI. Just because generative AI gives you the ability to do something, businesses need to think hard about ‘should we do this’ and ‘is this aligned to our core values’?”
“Trust with customers can be eroded very quickly. Businesses must consider privacy rules, unfair bias, disinformation risks, deepfakes, social engineering and over reliance on AI as threats to the trust they currently have with consumers and/or threats to their businesses.”
"What the Midlands has done well is capitalising on the industries the regional economy was built on," says Aaron Baker, investor at BGF. "The automotive sector and its entire supply chain has seen exceptional tech innovation in the region – from batteries to electric vehicles – and that's because of the talent, experience, and rich heritage here. And this has evolved into global leading verticals in the sector, such as Leamington Spa's dominance in the gaming industry."
Privacy
+
Katie Simmonds
Managing Associate, Womble Bond Dickinson
“If you are using ChatGPT or other similar tools in a way that processes personal data, you will have to comply with regulations such as the AI Act in the EU, as well as any applicable underlying data protection regulations, such as the GDPR. The EU’s AI Act essentially says for high-risk use cases, organisations will have to justify that use and perform an impact assessment, weighing up the risk of harm of the AI against the benefits to individuals and society.
“The GDPR is focused on transparency, and using people’s data in ways that they would reasonably expect. Both the AI Act (once enacted) and the GDPR have very high financial penalties for breaches.”
"What the Midlands has done well is capitalising on the industries the regional economy was built on," says Aaron Baker, investor at BGF. "The automotive sector and its entire supply chain has seen exceptional tech innovation in the region – from batteries to electric vehicles – and that's because of the talent, experience, and rich heritage here. And this has evolved into global leading verticals in the sector, such as Leamington Spa's dominance in the gaming industry."
Where next?
Impact on jobs
+
Ian West
Partner, Head of Tech Alliances and Tech, Media and Telco Sector, KPMG in the UK
“Generative AI is still in its infancy and the possibilities are endless. Based on the demand for support we are seeing from clients, we don’t see interest waning anytime soon.
“Likely in the short term we’ll see further exciting new business models and possibly a negative impact on jobs – though perhaps neutral job impact in the medium term as new jobs are created working with gen AI.”
"What the Midlands has done well is capitalising on the industries the regional economy was built on," says Aaron Baker, investor at BGF. "The automotive sector and its entire supply chain has seen exceptional tech innovation in the region – from batteries to electric vehicles – and that's because of the talent, experience, and rich heritage here. And this has evolved into global leading verticals in the sector, such as Leamington Spa's dominance in the gaming industry."
Generative AI has the potential to reshape entire economies. As machines gain the ability to generate sophisticated content across various domains, new markets and industries may emerge that could contribute significantly to growth. Existing industries, such as medicine, law, accounting and others, could also boost revenue by improving operational efficiency.
Creative industries could witness a profound transformation as AI-generated works become highly sought-after and commercially viable, though issues regarding copyright protection of AI-generated works will need to be addressed. Recent market forecasts have projected the industry’s value to surpass $100bn over the next 10 years.
Despite a drop in funding for UK tech startups, the recently found wave of support for generative AI has encouraged cautious investors to make an exception for companies in the industry. According to Pitchbook data, generative AI companies globally raised $1.7bn in the first quarter of 2023.
While funding has increased for AI startups, these companies are in a very different macroeconomic environment from the previous three years. Many will need to demonstrate a faster route to profitability to investors, particularly as the hype dies down. Generative AI startups will therefore have to prioritise ensuring their technology finds product- market fit, solves specific customer needs and offers a clear plan for monetisation.
The future impact of generative AI will require the technology to be appropriately applied across industries looking to benefit from it. Most businesses have existing IT systems that, whether recently made or years old, are not fit to integrate generative AI tools.
For companies looking to benefit from these tools, investments will have to be made to make the integrations or rebuild legacy systems from the ground up. This could be challenging given the rate at which new services are being developed. Companies will therefore need to ensure they can flexibly integrate AI tools responsibly and smoothly
Establishing a regulatory framework for generative AI has become a key priority for many governments. Major powers, such as the UK, US and EU, are working with top industry figures to develop regulation that doesn’t inhibit the growth of the technology, but sufficiently addresses safety concerns.
There are many unknowns about what these regulations will look like. However, they will need to cover the issues of copyright protection, data privacy, and cybersecurity. The UK has said it is taking a “pro-innovation approach”, while other major governments have been more cautious. Italy, for example, briefly banned ChatGPT over regulatory concerns. As is always the case when regulating high-growth technologies, governments will need to be adaptable and move quickly.
Katie Simmonds
Managing Associate, Womble Bond Dickinson
“Where an organisation is using generative AI technologies, it is vital that someone has oversight of how the technology works and that there is some sort of verification of the output being produced.
“To this end, we are seeing an increasing number of organisations creating new senior roles to deal with the use of AI technologies – including the creation of new roles like chief AI officer. This trend is likely to continue as we see the increase of regulation of AI globally.”
"What the Midlands has done well is capitalising on the industries the regional economy was built on," says Aaron Baker, investor at BGF. "The automotive sector and its entire supply chain has seen exceptional tech innovation in the region – from batteries to electric vehicles – and that's because of the talent, experience, and rich heritage here. And this has evolved into global leading verticals in the sector, such as Leamington Spa's dominance in the gaming industry."
Thank you to our sponsors
KPMG LLP, a UK limited liability partnership, operates from 20 offices across the UK with approximately 17,000 partners and staff. KPMG is a global organisation of independent professional services firms providing audit, legal, tax and advisory services. It operates in 143 countries and territories with more than 265,000 partners and employees working in member firms around the world.
Technology is at the heart of what KPMG does and part of the very DNA of our business. That’s why we’ve invested in a powerful team of technologists; creating a step-change in the way we work. Delivering broader, deeper expertise to our clients faster than ever. Our connected solutions stretch across a range of specialisms; from digital consulting, cyber and risk management through to AI, data and analytics, digital design, software engineering, and financial modelling.
Each KPMG firm is a legally distinct and separate entity and describes itself as such. KPMG International Limited is a private English company limited by guarantee. KPMG International Limited and its related entities do not provide services to clients.
|
|
Head of technology, business & commercial banking Lloyds Bank
Womble Bond Dickinson (WBD) is a full-service international law firm sitting amongst the top 20 UK law firms. Our team has expertise in advising clients across eleven key sectors and we support a wide range of local, national and international clients of all sizes. What sets us apart is the combination of our unrivalled local knowledge and national and international expertise. As a ‘full-service law firm’ we are set up to do everything our clients need.
We have a dedicated team of technology lawyers who are experienced in the fast-changing technology law landscape. This team encourages and supports businesses and organisations of all sizes operating in the technology industry, from worldwide leaders in software, electronics and communications to emerging entrepreneurs, providing a range of services to allow them to survive and thrive.
With today's focus on AI and machine learning, our specialist AI practice allows us to help clients navigate challenges, provide innovative solutions and anticipate new issues and regulations surrounding AI. Our commitment to client service is unwavering, and that includes staying at the forefront of AI technology and employing it to deliver the best possible results for our clients.
|
|
Since 1765, we have been providing expert support and guidance to businesses of Britain. Supporting British businesses and the UK economy continues to be central to what we do.Whether you are an entrepreneur starting out or an established business, we can help you manage your business. Growing our lending, supporting with day-to-day business needs, and helping to protect businesses from the growing threat of financial fraud, are just some of the ways where we can support.
For many tech businesses, access to banking other sectors take for granted isn’t always available. Our knowledge of the sector means we can provide banking that’s crucial to managing your money from the get-go. Our Relationship Managers are experienced in the sector and keen to support further UK tech growth. We have a range of different products and services designed to meet your banking needs so that you can focus on growing your business in this fast-paced industry.
As you begin to scale up, you need a bank that grows with you. Whether it’s managing FX, deposit options or access to 3rd party support, we can connect you to the right people to maintain your trajectory.
|
|
