The field of artificial intelligence had a break-out year in 2023, with AI technology finding itself a niche in virtually every industry and digital application on the planet.
As we welcomed 2024, your humble scribes at Transparently.AI sat down with two bona fide AI pioneers - our founders Hamish Macalister and Mauro Sauco - to capture their thoughts on the year that was, and what we can look forward to this year.
What follows is an edited and amalgamated transcript of their comments. This exercise helped inform our CEO Hamish's appearance on the BBC just after Christmas, which you can watch here:
Courtesy: BBC
Everyone loves a good prognostication. Please enjoy our offering for AI in 2024.
GenAI went mainstream thanks to ChatGPT, Claude and the host of LLM-powered chatbots. Dall-E and GenAI art took off. Google Deepmind released its Gemini model to cap off the biggest year ever for AI. Now we see:
Consumers are using GenAI for anything from travel itineraries and meal plans to research and content creation.
Corporates in all industries (especially data-rich sectors) are exploring GenAI and applied AI solutions to extract efficiencies, cut costs and improve productivity
At the same time that the big players were making announcements on their latest releases, the open source community was surging ahead with myriad systems and tools, making it dramatically faster and easier to develop and prototype new AI/ML products.
We saw increasingly robust global conversations around regulation, prompted by the debate around AI safety. That bubbled into the mainstream consciousness with Sam Altman’s firing and rehiring at OpenAI.
What we see emerging in regulation is a divergence between a hard rules-based approach (as with the EU’s AI act) and a guidelines-based approach (as advocated for by Singapore and ASEAN, for example).
This naturally provokes conversation around equitable access, to ensure that no countries are left behind by overly onerous restrictions to a generational technology with the power to substantially upgrade economic output.
We definitely could, as more countries realise the opportunity that AI affords to advance their economies. Singapore is going all in with its National AI Strategy 2.0. We're obviously following this closely because it's right in our backyard. The NAIS 2.0 is a 15-point plan over 3-5 years to embed AI into the national consciousness and the economy.
More countries will follow suit as they realise the power AI has to supercharge efficiency and economic growth.
Of course, something has to fuel that growth. Last year, we saw growing debate around sustainable energy use to power AI and machine learning. Training a single model uses more electricity than 100 US homes consume in an entire year.
In 2022, Google reported that machine learning accounted for about 15% of its total energy use over the prior three years. Part of the EU’s AI act attempts to address sustainable development and robust protections for the environment.
The notion of AI has entered the consumer and business consciousness. Ironically, the form in which it has taken is only a tiny fraction of what is actually happening in the AI/ML world; e.g. big data analytics, predictions, and so forth, as opposed to LLMs, which have received so much attention.
Nonetheless, businesses will come under ever greater pressure to try to understand and incorporate developments. It will be a race. At the same time, concerns will grow regarding misuse, biases, inequality, etc. This will create ever greater pressure for various forms of regulation.
However, forming rules/regulations for such an extraordinarily rapidly growing technology will be extremely difficult.
This could be the year where models get trained to independently execute small tasks
What we can see happening, certainly in mainstream applications:
It will become as natural to query your favourite AI chatbot as it currently is to search the Internet.
More courses will pop up to train humans on how to effectively query AI chatbots for optimal results
More interactivity from your AI chatbot as the LLM training improves. ChatGPT and Claude already are showing signs of independently augmenting your query or prompt results with their own suggestions
As a consequence, there will be more discussion around the concept of “agency” in AI, the development of a system with a complex range of desires, beliefs, and deliberations that can fuel intentional and independent actions.
This could be the year where models get trained to independently execute small tasks, like book you a holiday that it’s designed for you, or do your grocery shopping independently.
We definitely foresee more instances of “applied AI," especially in the corporate world. This is already happening, but it bears reiterating that there will be more and more evidence of specific usages of AI to fulfil certain (and often labour intensive) functions that will augment productivity and free up human resources to be more productively used in functions that only humans can fill.
These will be focused on industries that collect and store data in large volumes, which can be harnessed for machine learning and signal recognition. Eg. manufacturing, logistics, supply chain, financial services, health care.
Many of the AI startups that we meet are in this applied AI space. Transparently.AI is in this space: Our system harnesses millions of publicly disclosed financial data points to detect accounting fraud and manipulation.
Despite the ubiquity of stories around AI, there is still considerable scepticism amongst many regarding its benefits and applications. We expect the next 18-24 months will see significant change here.
The notion of AI will become accepted as something that is here to stay, can bring significant benefits, has myriad applications and will become embedded in ever more aspects of our lives.
Companies will be under ever greater pressure to understand this space and employ it to derive revenue/cost/productivity benefits. There will be something of an arms race. There’s a limited number of people available with the skills to assist companies to achieve this. Competition for those skills can become intense.
If you’re in a repetitive and time-and-data-intensive job right now, where machine learning can do the task at scale, at a fraction of the time that it takes you, we can understand why you’d be concerned.
We would focus on the opportunities that arise from this generational technology.
Governments and the AI industry can take the lead here in terms of helping identify clearly which jobs are at risk and helping people there to reskill or upskill. Help them to understand how AI can make their lives and functions easier and more efficient.
For example, the Transparently.AI system will reduce the amount of time a forensic accountant spends analysing a company’s balance sheet to seconds from weeks or months, previously.
Does that mean the accountant loses their job? No, what it means is that they can analyse dozens of companies in the same amount of time it used to take them to study just one.
We would focus on the opportunities that arise from this generational technology. Our product is just one AI system that can generate millions, if not billions, of dollars in efficiencies in financial services, accounting and auditing.
Multiply that across the entire planet’s worth of applied AI systems all doing the same for their industries. Imagine the quantum of additional economic output that is generated.
Yes, if you consider the pace at which the technology is developing, which is miles faster than other transformative technologies in human history. It took years for the Internet to catch on or become ubiquitous.
Granted, the ubiquity of the Internet has made it tremendously easy for everyone to access or adopt AI technology. Access is one thing, but also the speed of innovation that is happening at organisations like OpenAI, Anthropic and Google is mind boggling.
One estimate from Precedence Research: The global AI market size was valued at US$454 billion in 2022 and is expected to reach US$258 trillion by 2032. That represents a compound annual growth rate of 19% from 2023 to 2032.
Also consider the depth and breadth of conversation and debate that is happening around AI. There are conversations happening around use cases, safety and regulation happening today that weren’t happening in this intensity 18 months ago. That will just accelerate.