Artificial Intelligence is evolving at a breathtaking pace. A few years ago, creations such as OpenAI’s ChatGPT or Google Bard were only limited to far-fetched discussions at tech roundtables. However, in the last few months, a lot has changed. Developments in AI technologies have been occurring at a break-neck speed leaving several individuals and organisations speechless in awe and dubious with apprehensions. Stanford University recently unveiled its 2023 AI Index Report. The AI Index is an initiative by the Stanford Institute for Human-Centred Artificial Intelligence (HAI) which is led by the AI Index Steering Committee. The report essentially tracks, collates, distills, and visualises data related to AI. According to the institution the report enables “decision-makers to take meaningful action to advance AI responsibly and ethically with humans in mind.” The surge in AI-powered technologies is evident, however, the report has made some alarming findings about the rapid proliferation of the technology. Interestingly the report comes at a time when billionaire Elon Musk, scientist Gary Marcus, and Apple co-founder Steve Wozniak among 1800 other dignitaries signed an open letter seeking a six-month halt on the development of systems that are more powerful than GPT-4. Here are some key takeaways from Stanford University’s 386-pages long 2023 AI Index Report: Bigger & faster AI models are getting smarter, bigger, and faster. As of now, AI models are getting a million times more data in training than a decade ago. The report states that until 2014, most significant machine learning models were released by academia, however, now they seem to have been overtaken by industry. In 2022, as many as 32 significant industry-produced machine learning models contrasted with the mere three produced by academia. The widening gap is attributed to the large amounts of data, computational prowess, and money required to build AI systems, something which industries are well-endowed with compared to nonprofit organisations and academia. Radical improvements Today, AI models are far more capable than they were two years ago and this is owing to the technical performance benchmarks across language or text, image, and video assessments. The report stated that AI consistently reported state-of-the-art results, however, a year-on-year improvement on many benchmarks showed marginal improvement. It also highlighted the release of new and more comprehensive benchmarking suites such as Big-bench and HELM. AI models need more resources Training large language models is a costly affair. The GPT-(2019) was trained at an estimated $50,000, while Google’s PaLM(2022) cost $8 million. Interestingly, in the last decade, for the first time, year-over-year private investment in AI decreased. According to the report, the Global AI private investment was $91.9 billion in 2022, which is a 26.7 per cent decrease compared to 2021. There has been a decrease in the total number of AI-related funding events and newly funded AI companies. The report says that the amount of private investment in AI in 2022 was 18 times higher than that of 2013. AI accelerates scientific progress The report states that AI models are helping in accelerating scientific progress. In 2022, they were used to aid hydrogen fusion, generate new antibodies, and improve the efficiency of matrix manipulation. More businesses root for AI The percentage of companies implementing AI has more than doubled since 2017. Also, companies that have integrated AI reported lower costs and higher revenue. “Organisations that have adopted AI report realising meaningful cost decreases and revenue increases.” Rise in demand for AI jobs & job satisfaction with copilot The report says that there has been a considerable increase in the demand for AI-related professional skills across the American industrial sector. The number of AI-related job postings has increased from 1.7 per cent in 2021 to 1.9 per cent in 2022. According to the report, employers in the US are increasingly looking forward to hiring talents with AI-related skills. More AI PhDs are opting for private sector jobs (65 per cent) compared to academia (28 per cent) or government (1 per cent). Besides, AI copilot tools are making people more productive. As many as 88 per cent of Github’s Copilot users reported an increase in their productivity, and 74 per cent said that they can now focus more on satisfying work. Rising misuse of AI Even as companies scrambled to develop newer AI technologies, there have been countless instances of misuse of these technologies. The report has cited the rising number of AI-powered deep fakes and security breaches. Overall, the number of AI incidents and controversies has increased 26 times since 2012. The incidents have been attributed to the greater use of AI technologies and the awareness surrounding their potential misuse. The report has also suggested that AI systems can have serious impacts on the environment. Policymakers keen on AI The report analysed legislative records of 127 countries, and the number of bills related to AI that were passed showed a phenomenal increase - from 1 in 2016 to 37 in 2022. Another analysis of parliamentary records from 81 countries revealed that mentions of AI in global legislative proceedings increased by 6.5 times compared to 2016.
