The wondrous world of artificial intelligence (AI) is in rapid acceleration and change is afoot, so we asked Croud’s AI enthusiasts for their thoughts…
OpenAI’s public launch of ChatGPT has taken the world – and people’s imagination – by storm. It’s capabilities are vast, from being able to hold a conversation, write a vegan meal plan and even help developers write code.
Put simply, ChatGPT is a large language model that reads the internet and then uses that information to handle questions. For this reason it’s being touted as a ‘Google killer’, and the fact it’s being heavily funded by Microsoft also adds to that narrative. It takes an incredible amount of power to drive these things and Microsoft are offering OpenAI significantly discounted cloud services to power it – although a paid offering is coming.
This has triggered ‘code red’ at Alphabet. Sundar Pichai called in Google founders Page and Brin, with Brin filing a code check driving a tonne of headlines despite it being a minor thing. That code red has probably been escalated further as Google’s demo of ‘Bard’ was picked apart by an astronomer removing a cool $100bn from Alphabet’s valuation.
Why is this sparking so much interest?
In part, it captured the public’s imagination because ChatGPT was pushed out for free. This feels like a step change from prior AI announcements which have been locked away in a research facility. Think about IBM’s ‘DeepBlue’ which beat chess champion Garry Kasparov.
Google/DeepMind went one further, beating the ‘Go’ board game champion, Lee Sedol, with AlphaGo. This is considered much harder because unlike Chess, Go has too many potential moves to be able to process, forcing the machine to make a best guess on the next step.
Then entered ChatGPT, which was publically available and spread like wildfire over the internet.
In the minds of some, it’s also re-igniting the ‘search wars’ which Google was deemed to have won after beating competition from Yahoo! and Microsoft’s Bing. Google’s lead in search has felt unassailable over the last decade, and AI chatbots feel like elements of that search service can be done better, or at least to the same quality, on platforms other than Google.
Leaks of how ChatGPT is being integrated into Bing have already appeared, suggesting it would sit as a separate service alongside Bing Search, not as an integrated system.
What does Croud think will happen?
Martin Reed – Product Strategy Director
Chatbots will force Google to innovate, but they won’t be a ‘Google killer’. Google has a real strength and depth in AI. In all the storm of ChatGPT people seem to forget a Google developer was sacked last year for blogging that he felt Google’s LaMDA was sentient.
The quality of conversation those leaks demonstrated looked at least as strong as ChatGPT – but it didn’t capture public imagination because it was quite a technical thing. I would very much expect Google to already be sitting on something equivalent to ChatGPT. The speed at which Bard appears to be being rolled out suggests they have indeed already sat on it.
Which leads to the question: why sit on it? This Financial Times article does a great job of explaining it.
Google has a market share north of 90%. It also suggests Search is considered to be a high margin business. If you have 90% market share in a high margin business, would you unleash something that might radically change your product? Or would you develop it, and find ways to use it without rocking the boat. I believe Google’s hand has been forced now.
I think it’s more realistic that Google faces larger threats from:
- The platform is allowing the next generation of users to find the content they want, faster – especially when it comes to things like recipes. On TikTok, they don’t need to read two pages of content on why this specific recipe reminds them of their grandma.
- Having their entire ad model scrutinised and broken down to allow for fairer business practices – especially across Display.
Privacy / volume of ads
- There’s no doubt the search result pages have become more about ads over the years. That’s putting off groups of people and may encourage them to explore other platforms like Bing or Neeva. But also, the rise in ad blocker popularity is also likely to be undermining Google’s Ads model – something Google are attempting to handle with new launches of Chromium.
Chris Ford – Head of CRO
If you boil a search engine down to a simple function of ‘ask a question/make a statement, get an accurate answer/response’ then AI Chatbots could quite simply be the ultimate innovation – if done correctly – for a positive user experience. However, they are completely different tools and arguably shouldn’t be too closely compared.
Consider the internet as a library. Going to the library, you have different ‘functions’ or ‘experiences’ to help you find what you’re looking for.
For example, you’re searching for a gardening book. There’s an entire section of gardening books already pre-categorised to a pretty useful and granular level, offered up to you as a dedicated section of the library – this means you can find the most useful book for your exact need. Having said that, you still have to discover which book is the right one for you out of the entire bookcase. That’s arguably Google Search, or at least the traditional Search Engine.
AI and Chatbots (could) come in like a knowledgeable librarian – they might (hopefully) know the best books to match your specific need and where they are, and based on your conversation and responses to prompts they can forward you more to a specific book, or smaller collection of books. This approach is more personalised, and in many ways results in a more useful outcome and friendlier experience of going to the library. You can find what you need quicker, and learn what you need faster.
These are, in essence, two very different experiences – traditional search is a tool for finding a collection of information. You have to know what you’re looking for and how to find it.
Chatbots, however, are more of a tool for receiving information. You can simply ask your question, or make your statement, and receive a comprehensive and ‘accurate’ answer. The trouble is, how will you know what’s right from wrong? And will people just trust the answers that it provides, even if it’s false?
If this week has taught us anything, it’s that AI isn’t there yet and it can’t be relied on for 100% accuracy – there seemingly needs to be a push back to more material to read for a more complete understanding. Who knows, that may lead to the monetisation of that model for traditional search engines to maintain their respective positions (or improve it).
Peter Eckersley – Head of Organic
The promise of AskJeeves, realised nearly 30 years later.
This is the classic innovator’s dilemma for Google. Search currently is exposed as a process which has more friction than necessary, but fixing this threatens Google’s core revenue stream.
Why has search got more friction than necessary?
Google prides itself on having the best search results in the world. It has incredible systems to assess the internet’s content for thousands of factors in order to determine which piece of content should rank number one for a user’s query, and then confidently displays it in position one. It then puts four other links ahead of that. Google’s ranking requirements have also had a negative impact on search experience, leading to bland and low utility content ranking which in turn has led to certain demographics relying more on Reddit or TikTok to find their answers.
Will Bard or ChatGPT be the death knell of traditional search?
There are going to be certain search queries in which a user wants to complete their own research and do more in-depth reading and discovery. What we can be sure of is that it’ll have a serious impact. Informational queries can now be answered directly by AI in traditional search results, as well as Q&A platforms. This will increase the instances of zero click searches to some degree. The roll-out will be crucial in how this plays out.
From Google’s blog post, we cannot currently see any links to sources that Bard is generating its answers from, again, reducing the amount of clicks to publishers. How will publishers respond? Will they continue to allow their content to be used for free – or demand recompensation for it? Likely the latter.
Will publishers be able to block Google from crawling and generating answers from their content? Will they be able to distinguish between Googlebot and Bardbot (not currently a thing), or will Webmaster’s be able to use directives at a page level to prevent them being used by Bard.
Adoption will also be key. People have become experts at searching over the past twenty years, using as little effort to get the results they needed to see. We search by keywords and provide very little context. It’s easy, and a behaviour particularly suited to mobile. ChatGPT and Bard want us to search in the way AskJeeves encouraged their users to search back in the nineties, typing out full questions (and sometimes adding a please and thank you).
Will we be bothered to change our search behaviour? The answer isn’t a binary yes or no. For some queries we will and for some we won’t. It’ll be fascinating to see how this plays out in the numbers.
Emily Griffiths – Head of PPC
Personally, I think all of this is smoke and mirrors.
Artificial intelligence isn’t a new concept. Our everyday lives are reliant on it – from using the estimated time in Google Maps, face recognition to unlock your phone, suggestions of videos on YouTube or TikTok, using Siri, Cortana or Google Assistant…. and that’s just scratching the surface. From a PPC specific view, Smart Bidding has been around since 2016 and even the fundamentals of Search on either Google or Bing are reliant on machine learning – which is an application of AI.
I think Microsoft is very clever in integrating ChatGPT within their Search – even if it’s more using the PR and name around it – but I think it would be foolish to think that Bard by Google is a knee jerk reaction.
I don’t think this will massively change user behaviour when it comes to the search engine results pages – it’s already been there to a degree with Google’s Knowledge Panel giving you answers to queries without needing to go to the website. For now, the output has been focussed from an organic point of view, with little understanding over how this affects paid search so far.
As a side note, the marketing and language used around ChatGPT is also interesting – it’s based on AI, not machine learning (machine learning being a subset of AI). It’s also the first chatbot to capture the world’s attention. The accessibility and hype around it has brought positive PR to those who were cautious, hesitant or even scared of it. There’s no talk of how machines could control the world, as the output of ChatGPT is very human-like, using everyday language. The challenge is how uncontrolled it has been – what’s stopping the use of ChatGPT for school essays or official documentation?
Kevin Ellen – Global Director of Web Experience & Organic Innovation
When we look at search, we can see a trend moving away from static pre-generated responses and results. This is to deal with the changing landscape of how people are searching, where it becomes more conversational (think about follow-up questions) as well as rich content (videos on YouTube, TikTok, etc). Traditional search engines will have to move in this direction, or else they’ll continue to lose out on share compared to other platforms, such as TikTok..
But what are the risks of using AI at scale for these purposes?
Without safeguarding it will not be able to differentiate between potential harmful/classified and high quality factual information.
What tools can be given to websites to avoid their content being assimilated in the massive pool of information?
Creators of AI will have to provide websites with the right tools to ensure their data is not misused or misinterpreted by a machine.
It’s an exciting time, given the sudden rapid acceleration of technology and huge investment into this space from various large companies – I think search may change quite significantly and these developments will pave the way for a more natural way of using search. I wouldn’t be surprised if this could be one of those moments that will define the years to come – if legal implications of deploying these technologies doesn’t cause significant ethical issues.
Andrew Sandoval – VP, Biddable Media
I’m biassed by my age and digital experience, but from my experiences with ChatGPT, search as we know it likely isn’t going anywhere anytime soon. I think early discussion of ChatGPT has oversimplified search, how we use it, and how Google makes money from it. My thoughts are as follows.
Where ChatGPT will win is where ‘search’ has answers. These are low volume (only 14% of Google searches are a question) and low value searches for Google from a direct revenue standpoint to begin with. If I need to know how often a newborn needs to eat, I’m not buying formula or food in the moment. Of course though, that data input is tremendously valuable to Google and its advertising customers – assuming privacy regulations or course – for further, extremely monetizable search/shopping moments.
This ‘answer’ application isn’t very interesting – we’ve had the search box, voice search, and ‘I’m feeling lucky’ button for many years. It also suffers the same drawback of all of those, and the reason I’m sceptical of any widespread change to the search experience.
People like to search. They don’t want one answer, they want many answers, and they like to pick their own. Even when they have the right answer on top, they still want a few other options.
AI’s real power is context, as many have said, and that’s why Google, Meta, Amazon, and Microsoft are far better positioned to win any kind of AI showdown than a new AI company. Google knows how we search, email, and watch YouTube. Meta knows our friends, what influencers we follow, and what content we share and enjoy. Microsoft knows our work, how we do it, and how we game. Amazon knows how we shop, etc. These companies also have the massive capital and cloud computing resources to make this bet.
Google should be concerned about having completely dropped the ball to this point. Google Assistant showed the ability for contextual knowledge and follow-on questions in conversations years ago. Why does search feel relatively dumb? Why isn’t it integrated more closely with Assistant? My family has made the same pancake recipe numerous times the past months and the Google Home Hub still can’t figure it out or show me the right answer first (or even stop showing it to me on my phone instead of the countertop device). Google has had AI features like chat summaries in GSuite for months that Microsoft is only now teasing for Teams via ChatGPT. Their problem feels distinctly like one of product and integration, not capabilities.
With divided opinion even amongst our experts, the impact that these developments in AI and chatbots will have on search remains an exciting mystery to behold. If you’re keen to chat to one of our team on any of the subjects raised and how AI could affect your business, drop us a line and we’ll be in touch.