What does Generative AI mean for the IT Channel?

The promise of Generative AI is broad, with the potential to increase both the speed and quality of output and decision making, it is likely to have profound impacts on our business lives and society in general. Much has been written on the potential for AI to replace knowledge workers the same way robots did to manufacturing workers in the last 30 years. The likely scenario is that many tasks will be automated reducing the number of workers required in that area. As a global society, we have always managed to use technology to create more opportunity even when certain jobs have been made redundant by it. We will all be hoping this trend continues as the use of AI proliferates.

The recent increase in interest in AI has resulted from the significant investments in Generative AI by OpenAI, Google and others. Open AI is currently on the 4th generation of its ChatGPT offering which was trained using human input and an estimated 45 terabytes of data (1) (figures for GPT 3 as no figures released for GTP 4). As the most advanced and complete Generative AI model today, we will discuss ChatGPT before looking at the competition.

The processing power required to train an AI model is enormous and as the models grow in sophistication, so do their appetite for resources. All the major Cloud Providers have voiced their plans for reducing the Carbon footprint of their extensive Data Centres (2). These initiatives are likely to be challenged by the scale of processing power required for these new AI models. ChatGPT-3 is estimated to have produced over 500 tonnes of CO2 (3) and ChatGPT 4 will be greater although OpenAI have not disclosed any numbers. The scale of Generative AI compute requirements is not limited to their Carbon Footprint, as this time a paid subscription to ChatGPT-4 comes with a limit to how many messages per day a customer can use (4), it’s been reported that Microsoft even has restricted its’ own development teams from using too much GPU processing time to leave capacity for OpenAI and ChatGPT (5).

There are clear risks to business and society including the use of deep fake videos, misuse of AI for hacking purposes and potential human job losses. The fact that these models can produce misleading or incorrect results known as hallucinations will become a significant problem if we learn to rely upon their output without the relevant checks and balances. We will require Governments and the Ai Industry to maintain vigilance and provide process and support to ensure that the inevitable rapid adoption of the technology doesn’t come with too big a sting in the tail.

From an IT Channel Partner perspective, the most relevant solution right now, will be that from OpenAI. Despite being an independent company, multi-billion-dollar investments from Microsoft (6) into OpenAI over the last few years to fund the processing power needed to train ChatGPT has given Microsoft a high degree of commercial control on how the solution is deployed. Given such large investments and the success of ChatGPT, it is no surprise to see Microsoft building the AI technology into its’ offerings including Azure, Microsoft 365, Security solutions and of course Bing. It’s early days for the Generative AI industry, however, given the speed of advance and potential benefits, we advise IT Channel Partners to start the process of exploration to identify the most imminent benefits for them and their customers.

A recent series of Partner interviews by ChannelWeb (7) identified Marketing as an area where many Partners are already exploring and finding benefit with AI. The consensus is that using Chat GPT to create content alongside your existing marketing output to augment and improve it is a worthwhile initiative. We would also add that with Microsoft cloud solutions so ubiquitous in the customer base, investing time to understand how Microsoft is embedding AI into its’ solutions is going to deliver payback in the quality of support a Channel Partner can provide.

In addition to OpenAI, there are other broad AI models available and in development in addition to more focussed, application-based solutions. The two most directly relevant competitors to ChatGPT are Bard from Google and the Titan and Bedrock solutions from Amazon (8) (9). Neither are as complete as ChatGPT but as the basis for these models is well understood and progress comes from investment and compute power, the playing field could level or tilt rapidly. The Chinese market has its own leaders with Ernie bot from Baidu and Tongyi Qianwen from Alibaba. Other western tech companies are also investing heavily in AI, Netflix uses AI extensively to manage its customers, content and platform and may decide to enter the generative AI race (10). Apple has been uncharacteristically quiet on the AI front, we can be sure it is not sitting idly by letting the opportunity pass it, they are actively recruiting for their Machine Learning and AI division (11) (12). Last but by no means least, IBM created Watson in 2011 when it won the US quiz show Jeopardy. This early success hasn’t paved the way for AI dominance for Big Blue who have had many challenges with Watson despite huge investment. Recent tests have shown Watson to have similar capabilities to ChatGPT and Bard so IBM may yet emerge as a key player in this market (13). In addition to the major players, smaller solutions with a more focussed approach are being built from scratch or via APIs with the larger platforms.