The highs and lows of generative AI

AI tools have proven useful in business for years - predominantly in making sense of large amounts of data. The coherence of data points is extremely important for mergers and acquisitions, fraud analytics, emissions targets, due diligence and other services that mnAi’s data platform provides insights for.

Industries such as private equity and venture capital prove the effectiveness of AI within their structure due to its ability to streamline back-office functions and expedite labour-intensive background research involved in deal-making.?

Early stage investor SignalFire recently secured $900m in fresh capital whilst using AI to collect data on the performance of the companies that it targets. Also, in 2014, the Hong Kong based company, Deep Knowledge Venture, quirkily appointed an AI software (Vital) as a board member - even granting it a vote on specific investment decisions.

The current hot topic is language-fronting AI tools. ChatGPT has burst onto the scene with its uncanny ability to mimic human writing and its effectiveness at responding to prompts. ‘This technological revolution is unstoppable’, as the CEO of OpenAI puts it, iterating that its utility should be embraced, rather than being feared by those lamenting the loss of ‘human jobs’. The tools are not here to take over - just as the calculator has not made every accountant redundant, but rather better at their job.

However, generative AI is not yet fully polished. Bard cost Google’s parent company, Alphabet $100bn worth of market value due to it producing inaccurate information. Their rival ChatGPT has also caused controversy with its censorship of public figures and unnerving answers to moral hypothetical questions.?

Generative AI is only as good as the data it receives. The problem with this is that the system will automatically learn and latch on to societal biases that are already ingrained in the information it is feeding from. It will not magically adhere to an up-to-date moral code, for example, as it purely depends upon the quality of the data that is used to train the system.

As the ‘intelligence’ part of the name suggests, AI is still learning, and early-stage mistakes in certain software should not discount the development of such technology. Efforts have been made to mitigate specific issues, such as placing importance on meritocratic attributes of candidates to raise diversity. However, even with vast investment AI systems can still be unpredictable - a lesson learnt by Microsoft and Google… and surely not the final lesson on the road to a brave new world.?