My contrarian view of this week’s AI arms race skirmish

My contrarian thoughts on OpenAI's (GPT-4o) and Google’s (many, many AI-related) releases this week…

I think these releases from both companies were just “bridging” releases, designed to:

1. Keep both companies in the public’s eye.
2. Keep both companies in the investor’s eye.
3. Announce new capabilities to set the table for much bigger capabilities to be released in the near-to-mid-term.

This week’s releases serve a purpose, but their true value is really just to get both companies (and us) through to their next big milestones.

What just happened?

My take: The updates this week from both companies were just a setup play for more significant developments in the near future. Indeed, much of the releases (particularly from Google) were “coming soon” promises that aren’t accessible yet.

OpenAI has a strategy of gradually introducing new features, preferring iterative releases to allow us, the commoners, to smoothly integrate AI advancements into society (and gradually update our understanding/acceptance of them).

Google has been in catch-up mode for the past 18 months after OpenAI left them with their LLMs around their ankles with the release of ChatGPT in November 2022. This week’s I/O Conference AI releases were the first flexes that felt like they now feel they’ve caught up and are perhaps ready to nudge ahead in some areas.

So, while the latest updates are impressive on their own, such as the new multimodal models, enlarged context windows of up to 2 million tokens (Google only), much improved voice capabilities, enhanced chatbot latencies, increased empathy, ability to connect to hard drives (like Google Drive), and the new desktop app coming soon for OpenAI, it appears to me that the REAL value of these capabilities are as building blocks for the upcoming release of GPT-5 and (probably) bigger Google LLMs later this year.

Both company's stated goals racing towards creating AI agents (AI capable of independent decision-making without explicit instructions). 

They have been explicit and bullish about that (as have many other top AI labs around the world). It is the “next big thing” coming in AI.

Looked at in that light, it feels to me that the recent releases are really just laying the groundwork for this next phase. I.e… the new capabilities are all just stepping stones for bigger things to come when GPT-5 and or Gemini 2.0 is released later this year, with agentic qualities.

Both upcoming models promise to be SIGNIFICANTLY more powerful than their predecessors, and in my opinion all the capabilities released this week by both OpenAI and Google can be seen as direct steps towards giving their AI agents arms, legs, eyes and ears to the world, to act autonomously and intelligently.

"I'm gonna knock your GPUs loose" 

The power of distribution (and its thirst for cash)  

What this week’s presentations did demonstrate was power. 

Google’s search engine, which is arguably the best business model ever invented, sheds a mind boggling amount of profit. That excess cash is being funneled directly at AI development. The company previewed new hardware, new AI integrations into consumer products, new chips, and new models.

Similarly, ChatGPT has the reigns of the fastest growing, and arguably most advanced piece of consumer software on the planet. With 100 million active weekly users, and (for now at least) its hands in the bottomless pockets of Microsoft, they are driving a runaway train, with no thought of braking.

The merits of these attitudes and gameplans is a discussion for another time (for example, Google has now shown clear intent to cannibalise its own cash cow; the search engine).

Meanwhile, although it’s clear that an AI arms race is in full effect, it is not yet clear what will work for whom, and where the profit will settle. 

Both companies on display this week though, through these release events, were looking to boost their capital reserves, to power the engine forward to keep creating bigger and better AI models. 

Make no mistake (as was so astutely pointed out by my colleague Holly Joint on LinkedIn this week)... a big part of this week’s moves by OpenAI and Google was to keep investors and users interested, ultimately so they keep writing cheques and buying subscriptions.

By the nature of their financial constraints, AI startups have to pick their bets about what they’re going to build, and keep a narrow gameplan. 

Behemoths like Google and OpenAI have the benefit of being able to bet on everything at once. And they will invest in as much as they can as they race towards being the first to deploy competent and powerful AI agents.

The most significant capabilities released this week

For what it’s worth, these are the things that were released this week by each company that I found most worthwhile. They are worthwhile when viewed as standalone releases (short term view), but even more so when viewed as steps towards AI agents (mid-to-long term view):

1. The desktop ChatGPT app: Able to sit on your desktop (initially only for Apple) and “watch” whatever you’re doing and offer best practice advice and guidance.
2. AI talking to AI: A cool preview, but obviously a preview of what agents will do; take individual roles/tasks, and collaborate with each other to achieve bigger goals.
3. Empathic, low-latency chatbot voice chat: Lightning fast, verbose and intelligent chatbots that can interpret vision and audio from the world around them were released by both tech giants this week.
4. Agents in search: Google wants to move us forward to an era where its search engine does all the work for you, and presents you the answer on a plate. 
5. Google in Android: Google is developing a virtual assistant that will be integrated at the core of the operating system.
6. Gemini 1.5 woven into everything: Google is threading its most powerful current model into a range of its products, from Google Sheets to Google Photos to Gmail.

Enjoy the skirmish, but don't forget about the war

So, enjoy the breathless hype that’s filling your X and LinkedIn feeds this week. Admittedly they’re exciting to watch. I’ve lost count of how many “game-changing ways OpenAI/Google just changed the game.” 

But, while you’re munching on the popcorn, keep in mind there’s a longer term goal here for both players.

They’re priming us for AI agents.

Likely to be with us by this time next year.

In the meantime they need to keep us as consumers interested and buying subscriptions, keep investors stumping up with barrows of money, all to keep the AI-building machine churning.  

For the moment, Google has the advantage of sitting on top of the biggest pile of gold since Scrooge McDuck’s, with a bigger distribution network than Escobar in his prime. 

OpenAI have 100 million active disciples who kneel at the ChatGPT altar each week.

Both are landing solid blows in these pre-war skirmishes, and are gathering their reserves for the bigger battles ahead.

Stay tuned. 

(Multi-modally).