Search 2.0: The AI Arms Race
How Generative AI and Prompting is changing the whole tech landscape
Who knew that 2023 would start with such focus on Generative AI and how this is changing the landscape for the big tech firms and in particular the business of search. It is fair to say that Open AI with its GPT models, including its latest version called chatGPT, has disrupted the whole technology industry. The race for Search 1.0 had Yahoo a firm leader in the early days but late entrant Google stole the march with its new algorithm PageRank. It now seems we have another race for Search 2.0. Currently this is a two horse race between the partnership between Microsoft and OpenAI and the work of Google & DeepMind.
There is so much to unpack with these recent developments, from the details of the partnership between Microsoft and OpenAI to the fact that the underlying method used in all these large language models was first published by Google.
While the big technology companies are moving very fast to establish marketshare or dominance, they are trying to over-look some of the problems and challenges with live versions of these ML models. Google’s live demo a few days ago which wiped over $100Bn from its valuation is at the sharp end of the potential impact, and it does seem slightly unfair, as the chatGPT as identical issues which appear to have been in the most overlooked (at least in terms of market valuation impact). More details of the “partnership” between Microsoft and OpenAI are also surfacing, and very much surprises me how favourable it is for Microsoft (but maybe that is the climate for venture funding at the moment). I have previously commented that I hoped OpenAI would stay independent, as I did when DeepMind was sold to Google. It is a shame that innovative AI firms become subsumed by the big tech firms. Maybe the massive costs involved in developing these LLMs means that these AI firms need to align with one of the big cloud/compute companies.
From Search 1.0 to Search 2.0
Moving away from the companies and competing approaches, it is plain to see that the business of search is evolving. Like many areas of our world, the impact of AI capabilities is raising the bar of consumer expectations. Ultimately what we are seeing is the addition of an AI based chatbot capability being added to the underlying knowledge base of a world-wide search database.
This will allow us to have a conversation as we research topics on the internet. Allowing our initial queries to be refined and personalised through the interactions of a conversational interface. This new ability will provide a much richer experience to users (even though it was possible to do much search filtering and customisation via the command line – but most users where unaware of such functionality).
A Big Challenge
While the race has now truly begun – there is still much to fix. The current AI models would be trained on a corpus of information (for chatGPT is it limited to 2021 information) and therefore would struggle to access more up to date information – this is one of the main tasks now for Microsoft and Google. Linking the AI to the search data and keeping the two synchronised.
This is a bigger problem, as it has been shown that the LLMs hallucinate, providing information and facts that are false. How can the output from these conversational AI technologies be checked and validated. This is putting the responsibility of fact checking to the tech firms and search portals. Whereas before they were simply redirecting you to the website that detailed the facts (and they had no responsibility for that). Essentially they are now becoming publishers and creators.
I watched the recent interview with Bill Gates, and he talked about how Microsoft never had to worry (or take responsibility) for any content written in MS Word. The Social Media platforms changed that, and have had to take some level of ownership and responsibility for what they allowed users to put on their platforms. It seems that now this is moving to the Search Portals now.
A Legal Perspective
The elephant in the room, is what might happen from a legal perspective. All of these models have been build from pulling all of our text available on the internet. Essentially we have all unwittingly contributed to these AI capabilities. Whenever companies make alot of money from algorithms built from public data it obviously raises many questions (and lawsuits). How the big tech firms deal with this will be both insightful and will potentially become test cases for all other AI firms building similar applications.
There has been limited focus on the ethical considerations around this too. No doubt that will also come more into focus once the initial hype has cooled down.
An Inflection Point – The Arms Race for AI
This could be a major milestone and inflection point for the development of AI. Potentially creating a significant AI divide between the few global companies that have AI domination of certain applications and capabilities and the rest of the AI ecosystem. At the moment it is unclear to determine if this will be damaging to the industry of AI longer-term or a major benefit to the sector.
We should also reflect on how the first wave of search created lots of opportunities and evolved the internet to be knowledge fabric that has empower human developments and achievements (and helped to accelerate these). Search 2.0 has the promise to do the same, in ways that seem difficult to predict or understand at this stage.
2023 – A Disruptive Year
The next 6 to 9 months are going to be very intriguing – seeing where both Microsoft and Google go with this. But also how the other big tech firms respond. We have not heard much from either Amazon, IBM or Facebook, but this maybe because search is not core to their businesses as much as it is for Microsoft or Google.
I just looking through some of my old blogs – and it looks like I had predicted in 2016 that improved search with AI chatbots would be coming – http://pardoe.ai/wp-content/uploads/2016/02/AI_Predictions_2016.pdf
Need our help now?
Get in touch, if you need more immediate support