ChatGPT – The Scary New AI Innovation That’s the Talk of the Tech Set

ChatGPT – Searching Out What it’s Bringing to Our Digital World

ChatGPT is now a real buzzword – ok, one word and an acronym – for those keeping pace with tech and search innovations. The AI technology enables scarily human-like interactions to happen between the tool and its user.

For a global audience though that’s already familiar with chatbots – and voice assistants like Alexa and Siri – that might sound like nothing new or exciting.

But ChatGPT is actually wowing the world with functionality way beyond what we ordinarily receive from bots and assistants.

Headline outputs speak of ChatGPT being able to conjure up email content. Write and fix code. Hold meaningful text conversations. And – welcome to the moral maze – compose essays and complete job applications. A bit more, then, than helping connect you to the correct human assistant at the correct department.

So in this Hookson blog we’ll look at how ChatGPT – and its emerging rivals – might impact on how we search online. We’ll look too at the potential implications for businesses and CX. And we’ll examine the biggest, boldest and most recurring prediction that has followed the unveiling of ChatGPT: that it is spelling the end of Google.

ChatGPT – the Bouncing AI Baby That’s Just Said its First Words

ChatGPT is shiny and new. Its creators, the San Francisco-based OpenAI,  launched the chatbot on 30 November 2022. The tool quickly gained traction, snagging a million users in its first week, and 100-million inside its opening two months. For context, TikTok was tik-tokking for nine months prior to reaching that milestone, whilst Instagram took a hardly instant two years.

ChatGPT Plus: From Free-to-All to Pay For Some

A free-to-all sign-up of course helped ChatGPT takeup. It’s worth mentioning though that this, somewhat predictably, has resulted in periods of non-availability as user demand overwhelms server supply.

Reacting to this, and understandably looking to monetize its R&D, OpenAI has introduced the subscription-based ChatGPT Plus. Unfettered access, as well as faster response times, will set you back $20 per month.

So How Does ChatGPT Work?

ChatGPT answers questions and creates content – think a list as diverse as social media posts and product descriptions, code and essays – in response to a user’s query. The app has been pre-trained on a huge dataset of information including website text, books and Wikipedia articles. All 300-billion words’ worth is there, silently powering ChatGPT.

LLMs: the ABC of ChatGPT

Beneath the bonnet of ChatGPT is found an AI technology known as Large Language Models (LLMs). These are algorithms that recognise, translate, summarise, predict and generate content that’s all founded on the knowledge those immense datasets provide.

Training – both manual and automatic – teaches the tool to offer real-world text-based responses based on what the models have seen and learned.

For example, one training process might ask the AI to fill in words deliberately deleted from a paragraph of text. Then, during training, the models are ‘rewarded’ for high-quality responses. Repeating this process results in the ability to generate useful, conversational text, and fine-tuning comes via people reviewing ChatGPT’s initial results.

But Relax: You’re Smarter Than ChatGPT

The technology, of course, doesn’t ‘know’ things the way a human does; it works by matching a relevant response to a prompt. This is perhaps best explained by ChatGPT itself. Answering the prompt ‘what are you?’, it responded:

I’m a large language model trained by OpenAI. In a more general sense, I’m a form of artificial intelligence that has been designed to process and generate human-like language. I’m not a human and I don’t have the ability to think or feel in the same way that a person does. I exist solely to provide information and answer questions to the best of my ability.

That said, the technology recently passed – with flying colours – the US Uniform Bar Exam. That’s the test attorneys take to prove the knowledge and skills required to practise law. ChatGPT aced a score in the top 10 percent of participants.

ChatGPT – Will it be Disrupting Your Sector?

It sounds like fun to ask ChatGPT to create a Shakespearean sonnet professing love for your iPhone. Or have it compose lyrics about your cat in the style of Bob Dylan. But that’s the contemporary equivalent of spelling out risqué words on a pocket calculator.

The big question is: is there a role for ChatGPT – and related innovations like Google’s soon-to-launch Bard chatbot and Snapchat’s My Ai – across business and commerce, searching and learning?

The answer: there absolutely is.

Soon, these digital products are likely to begin disrupting and improving any number of sectors. So here’s a snapshot of how the technology may be impacting business, communication and CX:

Enhancing CX and Automating Responses – ChatGPT technology within chatbots and virtual assistants can enrich customer experience. In fact, OpenAI is encouraging app providers to harness its tech within apps, and so deliver deeper and more meaningful interactions.

Already on board in addition to Snapchat are Instacart and Shopify.

Instacart’s users will be asking the app about ingredients, and for recipe tips. Driven by data from its 75,000+ stores, customers benefit from added value and a ramped-up CX.

For Shopify users, ChatGPT’s technology will provide personalised recommendations and help with search queries. Again, the enhanced experience will create reasons to engage.

So for CX, across businesses increasingly automating our interactions, ChatGPT’s technology is set to make a tangible impact.

Translating Text Of interest especially to global enterprises, ChatGPT can perform fast on-demand text translations. User customisation improves accuracy. For example, users can tell ChatGPT about the likes of regional differences in word meanings. And you can reveal the guise of the original text – a poem versus a legal document, say – to output a fine-tuned translation.

Assisting, Virtually The powerful innovation has the potential to complete daily business tasks, and so make more efficient use of time. Into the hands of AI can go chores like appointment scheduling, sending and responding to emails, and managing social media accounts.

Educating and Training ChatGPT has the ability to explain hard-to-understand concepts. It therefore has the potential to become a powerful educational and training tool. That could either develop as a standalone solution, or as a component of a mix that relies also on traditional teaching and training methods.

Coding and Debugging To be clear, ChatGPT can’t (yet) write sophisticated code. But it’s already doing a good job of generating suggestions for fixing coding errors. The thinking is that if ChatGPT can deal with what you might call boilerplate code, it’ll free up time for coders to concentrate on more sophisticated code.

Another potential is a requirement for coding professionals to check and maintain AI-generated code, then use their skills to apply this code to useful digital products.

Brainstorming ChatGPT, armed with that massive dataset from the internet, has a lot to say for itself. It’s therefore an ally in brainstorming sessions. And because the technology isn’t perfect – as Open AI admirably concedes – it might throw up some pretty lateral thinking. And lateral thinking, in Hookson’s experience, can result in awesome idea generation.

Summarising Text ChatGPT is able to summarise lengthy texts and articles. Anecdotally, these are hit and miss at the moment, with some summaries missing out key points or creating a different spin entirely. And that brings us neatly to…

A Look at Limitations – Does ChatGPT Ever Run Out of Words?

A product like ChatGPT – ambitious, disruptive and in its formative phase – can’t hope to be perfect from day one. Limitations exist, and users should be aware of these. So let’s look at where this impressive technology is not quite hitting the mark.

It Prefers Shorter Articles And right here, as you read your 1,100th word – and counting – don’t we all? But whilst ChatGPT is mostly great at summarising and bullet-pointing, right now it’s still learning to produce longer-form content that stays on brief. You wouldn’t, though, bet against it getting the hang of this soon.

It Can Throw up Inaccurate Responses Because ChatGPT lacks both common sense and the knowledge and experiences a person has it can deliver answers that don’t make sense. Users challenging the tool’s inaccurate response will receive an admission and apology – just in case you weren’t spooked enough already by its humanish behaviour.

It Doesn’t Get Sarcasm. Or Humour OK, we all know actual people guilty of this too, but it needs saying that ChatGPT won’t pick up on prompts that contain nuance. And that can result in irrelevant or inappropriate responses. So it’s worth thinking about how you interact with the innovation. Imagine you’re dealing with say a Customs official or tax inspector and that should keep you right.

Multi-tasking Isn’t on its CV. Yet Currently, ChatGPT prefers handling tasks one at a time. Ask it to do too much and it’ll have issues prioritising.

It’s Open to Bias It’s as simple and as complicated as this: the data used to train ChatGPT can contain prejudices and bias. And that means content generated by ChatGPT can carry these too.

It Doesn’t Know Everything Whilst ChatGPT sounds like a bit of a know-all, it’s not. Really specific prompts about niche topics can be too big an ask.

It’s Like 2022 Never Happened Connected to the above, ChatGPT currently holds internet information up to the end of 2021. It therefore displays limited knowledge of events and developments beyond this timeframe. Of surprise to users therefore was discovering that ChatGPT knew of Elon Musk’s acquisition of Twitter – an event from late 2022. The tool may have learned this, and other post-2021 knowledge, via the research sessions the innovation undergoes.

So, plenty of ‘must-do-betters’ for ChatGPT to think about.

These are understandable though, and it would be pretty churlish to mark down ChatGPT – still in its infancy – for issues that OpenAI acknowledge and will be working to resolve.

Indeed, March 2023 saw GPT-4, the latest version of the technology, unleashed for ChatGPT Plus subscribers. A new feature includes an ability to respond to images – so it can, for example, cook up a recipe from shots of ingredients.

Ramped up too is the number of words of text (25,000) the innovation can handle. The result: greater searching and analysis of heftier documents, plus longer conversations.

Despite these leaps forward, perhaps the kind of ironic thing is that ChatGPT’s limitations – the bias, the issues with multitasking and nuance – actually contribute to and enlarge its remarkable near-humanness.

ChatGPT – Will it Talk Google to Death? In a Word: No

Plenty of the techerati have been quick to dub ChatGPT ‘the Google Killer’, and to proclaim, in advance of the crime, that ‘it was the Large Language Model, in the server room, with the dataset’. But this smacks of wild fancy. At least for right now.

Google’s iron grip on how and where we search feels unshakeable. It remains the world’s most visited website and across February 2023 held more than 93% of the global market (Statcounter).

Google Looks Concerned as Bing Gets Serious About AI Search…

That said, Google is paying attention to ChatGPT – and to Microsoft’s $10 billion partnership with OpenAI. The intention here is to incorporate ChatGPT into Microsoft’s search engine Bing – hampered, for too long, by, well, being Bing and not being Google. And by global usage percentages that struggle to hit double figures.

…But Google Hits Back With Bard

Google though is on the case; as we mentioned, its response has been to announce that Bard, its own AI chatbot, is incoming. Right now it’s being road-tested in advance of a rollout. Plus, AI has been at the heart of the likes of Google Assistant and Google Translate for some years.

Further to this, and without getting too technical, ChatGPT itself kind of owes Google. After all, the architecture at its heart, GPT-3 – Generative Pre-trained Transformer 3, if you must know – first popped up in a 2017 Google research paper.

ChatGPT v Search Engines

There are key differences also between ChatGPT and Google, indeed any search engine.

1. For one, ChatGPT’s training relies on internet text against Google’s training via user behaviour and web pages.
2. Next, guiding ChatGPT’s results are context and probability. Google’s by popularity and relevance.
3. Added to this, ChatGPT offers one answer to your query. Google et al will provide answers in the hundreds. And that enables deeper research via greater information.

ChatGPT is ‘more human’ of course. Google and other search engines are aloof by comparison. But if you’re looking to see when a flight is landing, or you absolutely must immediately know the current temperature in Rome, Google and its peers remain your destination.

In summary, ChatGPT’s rapid adoption – and certainly that mega Microsoft deal – has gotten Google’s attention. But let’s be real, rumours of Google’s demise are being exaggerated.

Search Will Change, Though. And AI Will Lead the Way

But what’s also certain is that search is going to change. And AI will be that change. ChatGPT’s formidable functionality has rocked search. Its technology – and its peers’ response to that technology is already impacting sectors and industries.

For Hookson, it’ll be fascinating to see how this emphasis on AI-based search develops. We’ll be on the pulse also with how smart use of AI can empower businesses to connect ever more deeply with their audiences.

Looking to Make Meaningful, Memorable Connections With Your Customers?

Let’s speak about how Hookson’s powerful suite of digital products can help you make the difference.