It was less than 12 months ago that the Computer Group (as it was then) had a session looking at the emerging technology of Large Learning Models (LLM) and in particular Chat-GPT and Dall-E from Open AI. A report of that meeting appeared on this website. We also debated the topic in September, an experience neither of the proposers would willingly go through again!!!! So with so much having happened in such a short period of time, it certainly seemed right to revisit the topic and see what’s happened.
First we supply some background, some developments and some resources for you to follow-up, then some notes and content from three member contributions.
What do we mean by Artificial Intelligence. Well, nothing that’s really appeared in the past 15 months it would have to be said. Some articles for you to consider …
Race to AI: the origins of artificial intelligence, from Turing to ChatGPT – in this article from The Guardian the background to what is, and what is not, artificial intelligence is explored and whilst recognising that a step-change has occurred, we have not yet reached the “nirvana” of artificial intelligence.
A simple guide to help you understand AI – in this animated webpage from the BBC, you have the “Janet and John” introduction to artificial intelligence. That’s not meant as an insult in any way, it’s a very straightforward introduction to the subject and if you switch the animations on you will get a full-experience.
It is however the Royal Institution’s Christmas Series of Lectures fronted by Mike Wooldridge that perhaps best defines the scope, the possibilities and the future of artificial intelligence. I seriously recommend that you watch all three lectures on BBC’s iPlayer.
Much of the attention in the past year or so has been focussed on LLM and Chat-GPT. Which? attempts to answer the question – “is it safe to use” in this article, but comes up with more questions than answers, but it does a bit of explaining along the way, but actually the best way to find out what Chat-GPT can do, is to try it yourself, which you can do here, after signing-up for an account on the Open AI servers.
Some experiments and experiences. We kicked-off here with a member telling us about his experience of using Microsoft Copilot.
Microsoft has released several AI “Copilots” that can complete different tasks for different purposes.
- At a basic level Copilot can be accessed directly through copilot.microsoft.com where it acts as a kind of Bing-ChatGPT. It can answer questions, create text, jokes, poems, songs and images much as we’ve seen in Chat GPT. When providing information it will reference the source material. And also send you on a shopping spree if you wish!
- Copilot is now fully integrated into the Bing search engine and Edge browser. Open Edge from the Windows taskbar to bring up the browser. Copilot is the top icon in the vertical tabs on the right. Clicking on it will open a sidebar which looks much like the page we’ve seen above. A different tab will open the AI Image Creator. Copilot in Edge can work with web pages, for instance generating a summary of complex articles. It will work in chat, compose and insight modes. It offers three conversation styles for your own composition – creative, balanced or precise!
- Copilot in Windows is currently being rolled out in Windows 11 updates. It can help with a variety of tasks, both relating to your PC’s settings and to generative assistance. For example, Copilot in Windows can turn on your computer’s dark room, organize your windows, write an essay, help you shop, and more with a simple conversational prompt.
- Copilot for Microsoft 365 involves the integration of the Copilot AI assistant into Microsoft’s 365 productivity software family. For example the suite of Microsoft 365 applications, including Word, Excel, PowerPoint, Outlook, and Teams, will be infused with Copilot to optimize workflow, such as generating text and tables in Word, summarizing and drafting emails in Outlook, and generating Python visualizations in Excel. This is clearly a more advanced development of Copilot AI, intended for the business market, currently on offer for £30/month!
- There are further business related Copilots for Security, Service and Sales.
- All are under continuous development so will no doubt involve rapid change and improvement over time.
- Links to some further info are included below.
What is Microsoft Copilot? Here’s everything you need to know
Hands on with Microsoft Copilot in Windows 11, your latest AI assistant?
How to use Copilot to control Windows
Microsoft COPILOT review | Watch this first – YouTube video
Of all the commercial developments in the past year it’s Microsoft that’s led the way, but Google’s working hard to keep-up, first with it’s Bard model, and latterly with it’s just announced Gemini software.
Our next presentation led us through some of the features of Bard and this article indicates the differences between Bard and Chat-GPT. Iain was particularly struck by the way the system seemed to learn from one day to another. An inappropriate response to a question of taking your grandchildren somewhere in Cardiff, had been replaced by a much more acceptable answer. How does this happen? Has it learnt something? Or, is it’s previous answer become part of the resource that it uses to find responses. I guess we’ll never know!!
Since the meeting the presenter also shared with us in the Artificial Intelligence Forum on our Thought grazing website (open to registered Cardiff u3a members) a really interesting and potentially useful use of AI in interpreting scrolls recovered from the eruption of Vesuvius. There are many really useful uses of AI, as the Royal Institution lectures showed us. I aim to add Topics to that Forum over the next few weeks (and months) as they emerge.
We then had an illuminating presentation from another member, that I encourage you to look at.
It starts off with his interest in Family History and his ancestor’s death from cholera which prompted him to enquire about that epidemic.
He then asked Chat-GPT to write an essay (which he’d had to do in an earlier life) on the topic “Liberal and Utilitarian Educational Thinking in 19th Century Great Britain: A Comprehensive Analysis with Reference to Cardinal Newman”
He felt that the attempt was worth at least a B+
He then shared with us some images created by Image2Go.com not all of which totally hit the brief he’d requested.
In conclusion, I shared a few issues …
Volkswagen says it’s putting ChatGPT in its cars for ‘enriching conversations’ – really? The first recording of a Tesla driver wearing Apple’s Vision Pro headset has already been posted to the internet. Where’s the common sense, in common sense?
A frankly rather surprising Judges in England and Wales are given cautious approval to use AI in writing legal opinions.
A totally unsurprising Britain’s MI6 chief says his spies are using AI to disrupt flow of weapons to Russia.
3 reasons why AIs like Microsoft Copilot and ChatGPT will only get worse — and why it doesn’t have to be this way – a suggestion that as more and more AI-generated content floods the internet, and as more and more reliable content providers stop the “bots” from trawling their sites, that the quality of information on the web will decrease. AI eating itself is a phrase that might come to mind.
Finally a very thought-provoking article – AI Needs to Be Both Trusted and Trustworthy – a theme picked-up on our Signal messaging group with a reference to the fact that Chinese academics wary of their reputations have moved towards implementing new guidelines to protect the quality of their research, as well as protecting it from being trawled – which may (as I indicated above) be counter-productive.