Advances in tech and investor hype have led to a new wave of generative AI startups — and existing ones leveraging the new tech. But as companies and investors lap up these new advances, concerns over data security and ethics have emerged, forcing companies to weigh the capabilities and risks.
As questions cloud the sector, we sat down with the experts to learn more about how startups can access generative AI in a safe and sustainable way — and what it means for the tech industry.
What does generative AI offer startups?
GenAI offers massive opportunities to industries by allowing companies to create content in the form of text, image, video, audio or code.
Rob Ferguson, global head of AI and ML for startups at Amazon Web Services (AWS), says that just about any company of any size and sector will use generative AI in the near future — if they don’t already.
“We have some incredible stats: for example, last year, we had the first drug companies actually make a clinical trial using AI — 10 years ago we had zero, and now we have seven drug companies doing clinical trials with AI-based drugs,” he says. European companies applying AI to medical research have raised a total of $2bn in the last 10 years, according to data from Dealroom.
“On the other end, we have Runway ML, one of our customers who helps create movies using AI. So we have examples of pretty much every sector — other than maybe firefighting — that have some cool examples of how they use generative AI in their applications,” he adds.
Coran Darling, an associate at global law firm DLA Piper with a focus on AI, says that many AI tools have the potential to amplify the productivity and capabilities of employees. “This is particularly relevant to early-stage startups where resources can be limited and employees can be stretched over several roles,” he says.
Many AI tools have the potential to amplify the productivity and capabilities of employees
For Francesca Tabor, founder of Fashion AI, a generative AI-powered fashion startup, the real strength of the tech lies in its power to improve products and offerings while being completely customer-centric.
“There’s the opportunity to use AI to conceptualise new products, test the market and get feedback before going into production — and this could be anything from a restaurant using AI to come up with new recipes to a fashion brand coming up with a new line of clothing,” she says.
Risks of data security and bias
However, a survey of more than 500 senior IT leaders revealed that 33% feel that generative AI is “over-hyped”, with more than 70% expressing concerns that the technology brings the potential for data security risks and bias.
“Bias is a real thing that we have to talk about. As an industry, it is our responsibility to put guardrails in place to ensure that AI-based solutions brought to market are trustworthy, fair and safe. For example, tools like Amazon SageMaker can help developers to identify bias in their data set,” says Ferguson.
According to the Pew Research Centre, a majority of ethical AI experts acknowledge the difficulty of reaching consensus on ethics, with 32% saying that ethical principles focusing primarily on the public good will be employed in most AI systems by 2030. It also says that progress is being made as AI spreads and shows its value — and as the rollout of new AI is inevitable, so is the development of ethical strategies.
Organisations, whether big or small, are already subject to several obligations that require data to be sufficiently protected and accurate, and their technological infrastructure to be resilient
There are also legal and regulatory hurdles that come with a new space like generative AI. “At a training level, for example, the data used may give rise to intellectual property and data protection concerns where the rights of others may be infringed if not appropriately accounted for,” says Darling.
But the opportunities presented by the tech could make it worthwhile for companies to prime their tech infrastructure, data strategy, security and ethical guidelines to tap into it.
“Organisations, whether big or small, are already subject to several obligations that require data to be sufficiently protected and accurate, and their technological infrastructure to be resilient. While generative AI may be a new vector, many of these existing obligations will also apply to its implementation,” says Darling.
Tabor says that the way to move forward is for people who are concerned about bias to actually start up companies that create the training data. "If you were to look at the majority of advertising in the world, that's not necessarily representative of diversity, it's a representation of our societal ideals. Unfortunately, all of those things do get baked in.”
How deep is your digital literacy?
Reports show that 66% of IT leaders fear that employees don't have the necessary skills to leverage generative AI successfully and 65% can’t justify the implementation of it at the moment.
“Understanding AI and machine learning, it's a very specific skill. A lot of people learn these things outside of their university or outside of mainstream employment. So you'll have to find entrepreneurial employees who are willing to learn these things,” says Tabor.
She adds that apart from training employees on AI, hackathons where you bring people together across different departments and teams can also be useful.
Darling says that depending on the extent of its integration and its uses, generative AI is likely to require investment of time and money to make the most use of it. “An organisation-wide approach to embracing AI will help unlock the full extent of its value, from buy-in from senior leadership, to exploring what new skills some companies might need to embrace, such as expertise in data science."
Trying out generative AI
But for Ferguson, integration is inevitable.
“You’re going to come in contact with AI whether you like it or not — for example, in my house, I use [Amazon’s] Alexa to control just about everything, so I use a nice chatbot to tell me the weather in the morning.”
Ferguson says that all companies should try generative AI as it’s easy to get started and the cost involved is low. For example, LightOn, a French startup that uses AI to help companies improve productivity, uses AWS SageMaker Jumpstart — Amazon’s ML hub that aims to help users develop ML models.
“Then as companies want to go further with the tech — or reach the middle layer of the stack — people can use platforms like SageMaker to finetune models. That’s when you're ready to really integrate it at a large scale for your business — and it's definitely not too early to get going with that,” he adds. Amazon SageMaker is a cloud machine-learning platform that allows developers to create, train and deploy machine-learning models in the cloud.
Startups are the lifeblood of innovation. We want to support them in their journey to developing incredible solutions with generative AI
There’s also AWS’s recently announced Generative AI Accelerator. Over 10 weeks, generative AI startup founders will be matched with business and technical mentors, network with industry experts and have demo sessions with potential investors and customers.
“Startups are the lifeblood of innovation,” says Howard Wright, VP and global head of AWS Startups, in a blog about the programme. “We want to support them in their journey to developing incredible solutions with generative AI.”
Amazon Web Services Inc. (AWS)’s new Generative AI Accelerator is a program focused on helping early-stage startups in the generative AI space solve big challenges to scale and grow. Applications are open until April 17.
