On 12th June 2017 Google Brain (the AI part of Google’s business) published a research paper Attention is all you need.
This was the invention of the Transformer which is the “T” in GPT (Generative Pretrained Transformer). In the paper they had successfully created the most successful method of an English to German translation task. Within a year it had deployed the technology to Google Translate to make it’s translations more accurate.
This would be a great example of ‘sustaining’ innovation. Used by incumbent companies to make their existing products or services better.
It would take Google nearly 6 years to release it’s own GPT model to the public and when it did, it was playing catchup with a much inferior product to Open AI.
On the other hand, computer scientists started creating more and more powerful GPT models. Open AI in particular moved forward at a pace with Chat GPT being released to the public in November 2022 and becoming the fastest growing product in history.
GPT-1 (2018): Introduced the idea of a “Generative Pretrained Transformer”.
GPT-2 (2019): Gained massive attention for its language generation abilities but was initially withheld due to “misuse potential”.
GPT-3 (2020): A huge leap with 175 billion parameters, powering tools like ChatGPT.
Why didn’t they innovate?
In one of my earlier blog posts I talked about the idea from The Innovators Dilemma about sustaining innovation vs disruptive innovation.
Google had / has a $200bn search business so why would you create tools that take customers away from this? This is effectively cannibalising your own business.
At the time of writing the number of queries that Chat GPT is getting is 7% of the level of search queries that google is getting. The question is whether over a period of time this will get bigger and bigger.
How does this relate to us?
As AI agents start to be rolled out over this year, the argument will be that they are not as good as the best human. It is being argued that they don’t have the breadth of capabilities of a highly trained negotiator. They would be correct.
Three questions to consider.
- Will they get better?
- How much do they cost?
- What will happen to the market when key companies master their use?
My suggestions
- Yes and very quickly
- 3% of the cost of person
- Improved service for a fraction of the cost, leading to very good, low cost operators.
If you enjoy good market share, good profits and a loyal and much loved team this change may be harder than if you are building from scratch.
New entrants to the market have the luxury of not having to consider the ‘downsides’ of what they might lose, as everything is a win.
I do see a genuine appetite amongst the Greenhouse OS community of agents to embrace AI Agents in the context of prospecting. I wonder if that will continue when we get into use cases that will involve less team members being required.
Changing when you don’t need to
“If it ain’t broke, don’t fix it” will be the cry of a many a successful business. The scale of the change in front of us is so large, that change will be required. The bigger the business, the bigger the change required. If we wait until it’s obvious, it may be too late.
Leave a Reply