Following the news that Microsoft will soon be integrating Copilot, an AI large language model (LLM) like the one behind ChatGPT, into Microsoft 365, we look at how this will help businesses, what OpenAI’s GPT-4 will bring, and Google’s response.

Copilot 

Microsoft announced last week that it will soon be adding Copilot to Microsoft365. Copilot is an AI large language model (LLM) and natural language conversational chatbot that can give the same human-like responses to questions as OpenAI’s Chat GPT can. Once integrated, Microsoft 365 (there’s no news of when that will be yet) users will be able to type instructions into (and ask questions via) the ‘Business chat’ text field – like the text field in ChatGPT

Microsoft says that Copilot will be integrated in the most popular Microsoft 365 apps – Word, Excel, PowerPoint, Outlook, Teams. This will mean that, with the use of normal language typed text prompts, be able to carry out tasks like:

– Writing/re-writing, sourcing, and editing documents in Word, just as you might write a piece using ChatGPT, thereby saving hours.

– Use Copilot to make meetings in Teams more productive e.g., by asking Copilot to summarise the key discussion points e.g., who said what, where people are aligned and where they disagree and suggest action items, all in real time during a meeting. Copilot will also be able to recap meetings and send you the notes afterwards.

– Creating whole PowerPoint presentations from a simple text prompt and adding any relevant content from other documents.

– Analysing trends and create summaries and graphs of data in Excel.

– Making it much faster for users to clear their inbox in Outlook, e.g. by drafting emails  and analysing long email threads in seconds.

– In short, Microsoft sees Copilot as major value-adding USP for its Office suite and is highlighting the time-saving, productivity-potential for users.

ChatGPT4 

Just days prior to Microsoft’s announcement about Copilot, OpenAI, ChatGTP’s creators and a close working partner of Microsoft announced the introduction of GTP-4, an improved and upgraded version of ChatGPT. Open AI says that GTP-4 is “OpenAI’s most advanced system, producing safer and more useful responses” and it can “solve difficult problems with greater accuracy, thanks to its broader general knowledge and problem solving abilities”. 

Images As Inputs Too 

As well as text inputs, GTP-4 can accept images as inputs and can generate captions, classifications, and analysis. For example, on the OpenAI website it gives the example of a GTP-4 user uploading a photo of some cooking ingredients accompanied by the question “What can I make with these ingredients?”.  The example reply for GTP-4 is a list of dishes that can be made with the ingredients featured in the image.

OpenAI also says that GTP-4, which was trained on Microsoft Azure AI supercomputers surpasses the earlier version of ChatGPT in its advanced reasoning capabilities and outperforms the earlier version by scoring in higher approximate percentiles among test-takers (90th compared to ChatGPT’s 10th on the Uniform Bar Exam).

Safer 

OpenAI also says that it has spent 6 months making GPT-4 safer and more aligned. For example, OpenAI says that, in its own evaluations, GPT-4 is 82 per cent less likely to respond to requests for disallowed content and 40 per cent more likely to produce factual responses than GPT-3.5.

Already Included In Other Products 

GTP-4, which is already incorporated in Microsoft’s Bing search engine, is also already being used in new products in collaboration with other companies and organisations, e.g. Stripe, Duolingo, and even the Government of Iceland (to help preserve its language).

Limitations 

Although ChatGPT has been extremely popular, and GTP-4 looks like being even more so, OpenAI itself and other tech commentators have warned of some important limitations and potential problems that businesses need to bear in mind about these AI models. For example:

– They can produce wrong and/or inaccurate answers and can potentially spread disinformation.

– The free version of ChatGTP is sometimes busy, i.e. it’s not available to all users in periods of high demand.

– GPT-4 can display social biases, hallucinations, and create adversarial prompts.

– Microsoft may not have server hardware needed to run the AI for Copilot across Office 365 and for GTP-4 in Bing (a shortage of GPU power). This could impact smaller businesses as Microsoft may prioritise capacity to bigger customers.

Google’s Answer?

Following OpenAI’s recent announcement of the general release of API access to ChatGPT and ‘Whisper’, its automatic speech recognition (ASR) AI model, Google has now announced that it’s giving API-level access to its LLM model PaLM to enable developers to build it into their apps and workflows. Google says that developers and businesses can now try its APIs and products to start building Google’s AI models through Google Cloud and a new prototyping environment called MakerSuite. Also, Google says that it’s introducing new AI features through Workspace. Google has also said that it may add PaLM’s AI capabilities to Google Docs, Gmail, and other parts of its suite.

No More Heard About Bard 

There have been no new announcements, however, about when Google’s direct conversational chatbot competitor to ChatGPT ‘Bard’ will actually be released. The original announcement of its impending release in just a few weeks, as an answer to the announcemnet of ChatGPT’s release, was reportedly met with criticism by Google employees for being “rushed” and “botched.”

What Does This Mean For Your Business? 

Although there were already chatbots out there, the release of ChatGTP and its game-changing success was really just the opening of the AI chatbot revolution which looks likely to be the next big leap in computing and business. As Microsoft CEO Satya Nadella puts it (with reference to Copilot) “we are at the start of a new era of computing” and as Microsoft’s Jared Spataro says, “Copilot marks a new era of computing that will fundamentally transform the way we work.” AI language models are now being released thick and fast, both through APIs so developers can make new, much more powerful apps, embedded in search engines and across productivity suites (Google and Microsoft), and as online chatbots for all end users. ChatGPT is already onto its next big chatbot upgrade with GPT-4. For businesses, all this will mean increased leverage and productivity, time and cost savings, better business insights, and being able to harness new value adding powers that can transform work and outcomes. We are still at the beginning of an AI revolution that holds a lot of promise for businesses going forward.