Updated June 11, 2023
Since I first wrote about the amazing AI developments with ChatGPT in February a great many more advances have been made. I summarise some of the more important here.
Of the mega IT companies Microsoft was the first to act decisively as they have been working with OpenAI, the creators of ChatGPT, for a number of years, and bought a 49% stake in the company. In March 2023 announced Microsoft 365 Copilot which will integrate this new era of generative AI into all Microsoft software products providing extra assistance to the end user in a safe manner.
Azure is the commercial cloud service offered by Microsoft. At their annual developers conference Build just weeks ago in May, Microsoft announced the Azure OpenAI Service which allows customers to use the powerful conversational AI model in their own applications and services. Microsoft themselves used Azure OpenAI to generate about a dozen specific copilots for their main software products. Consumers will reap great benefit from Windows Copilot, which leverages ChatGPT to provide users with personalized and contextual assistance across Windows 11 and Edge (Microsoft clone of Google Chrome). But the more important to all users is Bing with ChatGPT which brings Bing search, and hence up to date information, to ChatGPT whose model data is only current to a year or so ago. On any device (phone, tablet, laptop) in an internet browser go to bing.com/chat.
Of course at Christmas Google realised the huge success of ChatGPT would form a huge competitor to Google search and its lucrative advertising income. Google have been developing and using AI for a number of years but did not have a consumer AI product to announce. Over the last months they have put in a huge effort to come up with a ChatGPT competitor in Bard. Google IO is their annual developers conference where they announced a new Pixel phone, an impressive Pixel tablet/home speaker combo and their first folding phone. In AI they ended the Bard waitlist and made it available across 180 countries. Go to https://bard.google.com/ to try it from any browser on any device.
At their developers conference Google also announced a number of generative AI enhancements to their competitor apps for Microsoft Office. In particular Gmail is able to generate intelligent replies, produce summaries and various other email composition features. Docs which is a word proceesor now is able to brainstorm ideas, proofread whole messages, and compose and rewrite message content. Google’s PowerPoint equivalent Slides gains the ability to auto-generate slide shows with images and videos included. The Google spreadsheet Sheets can generate data insights and analysis from raw data tables. If you use Gmail you will increasingly become aware of these enhancements over time as they are introduced.
Google stresses the fact that Bard is experimental and that features are being introduced gradually in a safe and secure fashion. However, as you might expect, this approach has been adopted by OpenAI, Microsoft and all other companies who are participating in the generative AI advances. From my almost 60 years experience in information technology it is no exageration to call this modern AI revolutionary; it is up there with personal computing, the internet, world wide web, Apple iPhone and apps, the cloud, and social media as life-changing IT advances.
While generative AI has already demonstrated benefits in many diverse applications and industries, it also poses some ethical and social risks such as security, bias, privacy and misinformation. The bold generation of incorrect information by conversational AI apps like ChatGPT and others is referred to as hallucinations. Of course the companies behind these apps, including open source software apps, are constantly striving to reduce hallucinations and the other risks. Now when you use these apps they will frequently refuse to answer as the testing and feedback by paid testers and the general public bring problems to light.
The threat posed by these risks has led to several organisations to call for a pause or restraint in the further development of generative AI. Many of the companies competing in this field of course are reluctant to reduce their pace of development. They make the important point that bad actors like criminal organisations and some governments will ignore calls for restraint, and the good actors must make advances in generative AI to counter the negative effects. However virtually all AI companies are calling on governments to introduce guardrail policies to define the boundaries of good and bad use of AI.
In 2018 the European Commission’s data privacy regulation called the General Data Protection Regulation (GDPR) came into effect and has essentially become the world standard. Once again the Commission has led the way in AI when in 2021 they put forward a proposal to lay down harmonised rules for the use of AI - the AI Act.
In brief the AI Act addresses specific risks of AI apps, sets requirements for AI apps for high-risk applications, defines obligations for AI app providers and users of those apps and attempts to address enforcement provisions.
I am glad to say that Australia is also acting on the need to understand and potentially control AI. In February 2023 the federal government asked the National Science and Technology Council to report on:
This report was delivered in March 2023, and now the government has put out a discussion paper for us all to provide input by 26 July 2023 on the governance measures to be put in place. Links to the report (30 pages) and the request for online submission appear on the Chief Scientist site.
The era of generative AI technology will envelop us all in a few short years. It is up to every one of us to at least become aware of the basics of this new technology in order to reap the benefits and avoid the pitfalls.