GenAI, what new challenges will emerge from this revolution?

Xavier Lazarus
Elaia
Published in
3 min readSep 4, 2023
Photo from Google DeepMind

Following the first chapter, “GenAI, a tsunami for the tech world” published on 1st August, here is the second chapter.

This sea change will see two major issues arise for product and research teams.

The first challenge is an economic one, and it could prove to be a major one: we need to move from a model where knowledge gains meant costly R&D investment (experienced research teams have seen their costs multiply in recent years with the rise of tech start-ups), but not very expensive in terms of operating costs (generally hosting costs), to a model where the cost of entry will be lower, borne by major AI players, but based on a cost model linked to usage. If we take this model a step further, we are moving from research financed by large Capex to less resource-intensive research that will involve bigger Opex throughout production. However, this per-use pricing (based on API call requests) presents a significant risk of margin loss and should raise questions on the areas in which LLMs should be used, and above all how to optimise their use within companies’ tech stacks. We need to prevent software publishers’ margins from being lost to GenAI companies, in the same way that e-tailers’ margins have been decimated by large online advertising platforms.

The second issue is one of sovereignty and strategic independence. We could be facing a turning point in the history of technology, with the entire software market becoming dependent on a handful of major technology players. The ongoing revolution in generative AI, where the most widespread models currently in use, such as OpenAI, resemble centralised black boxes controlled by major tech players, is currently heading in this direction. Nevertheless, and probably in light of the experience that has led to the domination of GAFAM in B2C, increasingly tech players are now emerging in favour of more open-source models, which would make it possible not only to avoid over-dependence, but also to retain control over the management of the data and usage of LLMs. One example is Mistral.ai, a company that made the headlines a few weeks ago. Founded by former Google and Meta employees, Mistral.aiaims to offer an open, distributed model that will give back control to user companies and avoid the notorious centralising effect.

Generative AI is set to revolutionise the world of tech, in particular by making computer development capabilities accessible to as many people as possible and structurally changing certain markets. We are at the beginning of a new era. If we draw parallels with the Internet and the mobile Internet, which enabled the rise of SaaS models, already at the root of recent major revolutions in B2B, it is difficult to predict precisely what the broadest effects will be on our ecosystems. One thing is certain: if we don’t fully embrace this paradigm shift, we not only run the risk of being left behind by progress, we also run the risk of becoming irrelevant in the future.

You can also find the 🇫🇷 version of this paper in Maddyness here.

--

--