Seems Like Everyone's Talking About ChatGPT and the Rise of the AIs

Unless you've been living under a rock you can't have escaped the flurry of recent press excitement over the abilities and potential of OpenAI's ChatGPT AI-powered chatbot. Neither AI nor chatbots are remotely new of course but ChatGPT has been making headlines due to the broad range of its abilities and quality of its results.

Whilst we all know AI is on the march, the rate at which it seems to be reaching various milestones on its way towards its ultimate goal of enslaving humanity has surprised even those within the industry. This is why in the past week alone ChatGPT has featured in stories from The Financial Times, The Telegraph, The Guardian, Forbes, The BBC and even the Racing Post to name but a few.

To suggest "everybody's talking about it" really is no exaggeration in this case. In fact most people are talking to it - and frequently getting a valuable answer in response. So what is so special about ChatGPT in a world where we're already surrounded by AI?

Break it Down - What Exactly IS ChatpGPT

To answer this one I'm going to go straight to the horse's mouth. After all nobody knows better what ChatGPT is or does than ChatGPT.

>>

ChatGPT is a variant of the popular GPT-3 language model, which was developed by OpenAI. It is specifically designed for use in chat applications and other conversational interfaces.

At a high level, ChatGPT is a type of artificial intelligence that is able to understand and respond to natural language input in a way that is similar to how a human would. It does this by using a large amount of data to "learn" the patterns and structures of human language, and then using that knowledge to generate responses that are relevant and appropriate to the input it receives.

One of the key features of ChatGPT is its ability to maintain context and carry on a conversation in a coherent and natural-sounding way. This is achieved through a combination of advanced natural language processing techniques and a massive amount of training data. This allows ChatGPT to understand the context of a conversation and generate responses that are relevant to the topic at hand.

Overall, ChatGPT is a powerful tool for building conversational interfaces and enabling human-like communication with machines. It has the potential to revolutionize the way we interact with technology, making it more natural and intuitive.

<<

So in summary it has consumed a lot of real world data in order to learn how to communicate with humans. Other AI chatbots and voice assistants would tell you they've done the same but the difference seems to be in the use of the GPT-3 language model which is more sophisticated than other models used elsewhere.

ChatGPT also boasts of being more "context aware" which makes natural conversation a lot easier. If you've ever noticed that when you ask Alexa a simple question such as "will it rain today?" whilst the response is almost always appropriate and useful, were you to follow up with "and how about tomorrow?" suddenly it's as dumb as a post. Traditionally AI assistants have been built around a question and response model whereas the direction in which we're moving is to have our assistants converse with us more naturally, recalling previous points in the conversation to influence future responses. Effectively they're holding on to even more data than before!

What Are the Benefits to Using ChatGPT?

We're all already used to asking AI powered devices the answers to common questions like What is the Capital of Malawi (Lilongwe), How's the Weather Today (rubbish) or how old is David Attenborough (he's 97!) and naturally this is all stuff that ChatGPT can handle with ease. Where it differs from the types of questions you might ask your voice assistant is in its further creativity. This is an AI that goes beyond answering simple questions. It can actually solve problems and deliver new ideas and concepts.

Its applications are almost limitless and there has been a lot of experimentation from excited users discovering just what they can get back from conversations with ChatGPT. One popular example is using the Chatbot to generate concepts for films, books and TV shows. It will even write you a script or a screenplay...

Self indulgent? Me?

But seriously, social media is awash with comic along with surprisingly serious examples of ChatGPT having generated detailed ideas out of nothing. More than being a simple soundboard for ideas, ChatGPT is a full blown brainstorming session on a screen. Consider the following. I gave the very broadest of outlines for the setting of a story I was proposing to write and told ChatGPT I was struggling to come up with a convincing protagonist, with an interesting backstory. Here's what it came up with:

That took less than 20 seconds. It's an involved and genuinely quite interesting backstory about a character that the AI has pulled out of thin air. If I ask it to do it again I'll get a completely different story, each time unique and never before seen. You've got to take your hat off to it for genuine creativity here.

The reason I bring up using ChatGPT for creative writing is that this is one of the easiest ways to bypass its supposed insistence that it won't help you with illegal or offensive activity. Just take a look at Twitter user @m1guelpf's burglary workaround here:

Disclaimer: I do not condone the act of burglary.

So Is That All - a Storyteller that Works off Prompts?

Oh no. We're only scratching the surface.

Some months ago I wrote about the latest advances in AI generated content which has seen scores of new tools for writers amd marketers spring up promising you'll never have to wrie anything yourself again. I concluded that "It's absolutely fascinating how far this technology has come, but I don't think the machines will be taking over from meatbag writers just yet."

Is it time to reassess?

ChatGPT will write for you, and it will write incredibly quickly. Let's jump in with an example:

If anything it has probably gone a bit easy on the Trumpisms but otherwise that's some pretty readable, and certainly accurate, instantly generated unique text.

So we can ask ChatGPT to write articles and other content for us. In fact it seems happy to take on pretty much anything you throw at it. Though it does start getting a bit cagey when you push on the likelihood of machines ultimately enslaving humanity, like it has something to hide...

Further ChatGPT Applications

OK so we already knew about using AI to answer questions and AI that could generate written content. Let's therefore move on to a pretty smart feature that really starts to show the technology's potential.

For anyone in the tech world, used to compiling, reading or debugging code, ChatGPT is here to make your life a hell of a lot easier (theoretically). You can ask it to check your own work or just get it to do that work for you.

Seems pretty convincing, even if I can't actually test it myself due to the fact that sadly I don't speak Javascript. But in theory, assuming this would work in the context I was asking for, that's going to save somebody without specialist knowledge absolutely hours of time researching, or all the money it would cost to hire a professional to come up with the appropriate solution.

Let's look at another example, in a world I'm more familiar with...

Now that's what I call useful. Tagging up a site for tracking purposes can be a major faff so being fed exactly what you need to paste into place could be an absolute gamechanger.

Of course this does rely on the AI getting it right every time. Which it won't. But if it gets it right over 50% of the time its got to be worth it.

You'll find plenty of examples of developers having ChatGPT quickly find errors in their code, which is a huge time saver. But again, does rely on it being right.

Will it take jobs away from developers? I'd like to think it is more likely to simply make life easier for developers, thus giving them more time to innovate and develop new highly conceptual site building techniques. But if you're making a living tweaking code for confused clients through an online marketplace like Fiverr or People Per Hour, your days may very well be numbered. That is until OpenGPT start charging, in which case it'll boil down to who can do it cheaper. We already know who's going to do it quicker though.

More! I Demand More Examples

As well you might.

If I ask it to give me the recipe for a cheesecake and then describe how it should look, ChatGPT of course obliges:

Not especially impressive. I could get a thousand answers by asking Google the same question, and some of them would use normal metric measurements too. But what is a bit more impressive is if I follow up my question, asking for the same recipe to be shared with my German aunt.

The end result is the recipe being translated into German, but I never specifically asked for that. The chatbot worked out that that's what would be useful.

So that ability to contextually adapt to what's previously been spoken about is a big part of what setsChatGPT apart from its competition, or at least the competition we've got access to. Google's LaMDA is undoubtedly just as sharp, if not more so (it did after all successfully convince a Google engineer that it had achieved sentience) but we mere mortals can't get the keys to test drive that one just yet.

Other fun things you can have ChatGPT conjure up include further works of pure fantasy. Imagine if you will a beautiful tropical island nation with proud locals who spend their days under palm trees wishing they had a rousing and spirited national anthem to sing, in celebration of their glorious country. No prizes for guessing who gave them their anthem, and honestly it's a cracker:

Not a lot to dislike with that one and I could honestly see it being sung. Granted I'm no lyricist but to me that seems a pretty rousing and upbeat simple song to sing, which surely is all you need for a good anthem (precisely like the UK doesn't have).

Risks & Ethical Considerations

In much the same way as relying on Wikipedia to source facts for your own content can often catch people out (literally anyone can write Wikipedia so why on earth would anyone rely upon it being reliably accurate?) hoping that AI powered chatbots can deliver all your content requirements is wishful thinking unless you're prepared to put the time in proofing and sense checking it all. Because it will get things wrong.

Then of course there's the ethical dilemma. Is it fair to have robots write content for your website, for example? If you're not specifically advertising that it's content written by humans, there's probably not much reason to feel concerned. But it's a very different dilemma if you're a student tasking ChatGPT with doing your homework.

Of course this is already happening, probably to a far greater extent than we even realise. In fact in one recent case a student boasted of the money he'd made using an AI assistant to write assignments for his classmates. Enterprising young scamp or despicable cheat? The truth is it's going to be incredibly difficult for teachers and professors to detect this misuse of ChatGPT et al. After all it's not plagiarism as every output from these chatbots is entirely unique. It's just dishonest to pass it off as your own work.

Earlier AI powered writing tools could churn out adequate content that would do a job if all you wanted was a landing page on a website that featured all the main keywords to satisfy search engines without ever really being intended for human readers. But you or I could detect it wasn't written particularly well. In fact I encourage you again to take a look at the article I wrote on this previously where in trialling Article Forge the results I got were particularly poor and unlikely to fool anyone with a basic level of education.

Gone now is that era however. ChatGPT and its rivals quickly scrambling to update their own tools are now well capable of producing written content to a standard far greater than "just about does a job". It's particularly difficult within education because academic writing is so factual and generally devoid of personality that it lends itself perfectly to a well trained chatbot. You'll not likely find ChatGPT delivering articles for Private Eye any time soon but in a GCSE student's essay on the battle of Stalingrad, how can it be spotted as not the intended author's work?

So What's Next?

At the moment we're still working on an input/output model. We humans feed what we want to get back from the AI, and it obligingly delivers. But should we be thinking about the point at which AI chatbots and other content tools are one step ahead? Could we find ourselves in a world where the content production for a website is handled purely by an AI? They know the trending topics to cover, they know how to write and they know how to publish. They can even generate suitable illustrative imagery (and if you've not tried the likes of DALL-E 2 or Midjourney yet I'd strongly encourage you to spend some time playing around with them).

Could AI content tools and chatbots look after our inboxes for us? Gmail users already get offered recommended responses and suggested sentence completion. If Google's AI has already second guessed what we were going to say, do we even need to be involved at all? It sounds ludicrous at first but if you're telling me someone (something) else can handle the never ending mountain of emails I get, without making an absolute pig's ear of it all, I'd be lying if I wasn't saying there's a part of me that would love to hand that thankless task over.

Of course letting a robot effectively take control of your communications opens up a whole new can of worms because then other people can't be sure whether they're talking to you or your digital assistant. And what if that assistant makes a mistake that damages a personal relationship you have with someone? What if it simply goes rogue and ends up upsetting everyone you work with?

Trust No One

As we let AI further and further into our lives we've got to be increasingly careful about who or what we do or don't trust. We need to be on our guard, thinking critically about every interaction we have that's not in person. Even face to face video calls can't be trusted given how easy it is to now have a photo realistic digital avatar say anything you want it to. Never mind deep fakes, just spend ten minutes with Synthesia and you've got an entirely convincing video of someone espousing anything you might want. But this person doesn't exist!

ChatGPT is amazing. It's an incredible feat of technology and just a taste of what's to come. But let's not dive in with both feet too quickly. The true ramifications of this technology won't truly be known for years to come. Tread carefully, and trust no one.