OpenAI DevDay - "The Robots are Coming!"

I have been soaking in all the information that came out of this week’s OpenAI DevDay in San Francisco. It was exceptional and has made me realise that our AI advancements are accelerating faster than we are ready for. I thought at best we would see all this announce at best in 4 or 5 years. Nope it is all going to start hitting soon!

4 big things for me!

One. AI is getting much cheaper and much faster!

Two. Real-time APIs…wow!

Three. The full o1 model revealed

Four. This will hurt VCs and startups in the middle.

Then I have two insights for you to take away!

?1. Intelligence Is Getting Much Cheaper and Faster

The first thing was really suprising for me. OpenAI announced that intelligence is becoming not just smarter but also significantly cheaper and faster. They've increased rate limits on tokens to 10,000 per minute and slashed costs by 50%. WHAT?!

Then we saw Anthropic, always quick on the uptake, immediately matched these improvements. It struck me that we've never lived in a time where intelligence, artificial or otherwise, is advancing at such a breakneck pace while simultaneously becoming more accessible.

We have never lived in a world where intelligence is getting cheaper and smarter at the same time. This is genuinely Earth shattering.

We're venturing into completely new territory here. The barriers to entry for sophisticated AI applications are lowering, and the possibilities seem endless. I remembered thinking last night, "We're witnessing the pioneering true pioneering into new frontiers." The thought was both exciting but slightly unsettling. But let me tell you why…

2. Launch of a Real-Time API

Next came the launch of a real-time API. Before DevDay I was thinking that I should write an article on the impact on the API space because of GenAI. It is going to be completely different now! Things are moving quickly.

This isn’t ?just a technical upgrade they’ve announced; it’s a doorway to new possibilities. Think true voice-to-voice products could now become a reality and very quickly.

Problem statement: I appreciate some may not understand why this is important. There was a problem before. Here is an example to help: Imagine an AI managed autonomous vehicle that needs to process sensor data, make navigation decisions, and control mechanical components in real time:

• Before Real-time APIs:?The time it takes to process sensor, data might increase with the number of obstacles detected. In high-traffic scenarios, this could delay decision-making, leading to unsafe situations.
• With Real-time APIs:?The vehicle can process sensor data and make decisions in constant time, regardless how complex the environment it is driving in.

OpenAI have truly broken new ground here. I get it, it makes technical sense, that is all good, but I don’t think any of us really appreciate how much next year will change.

In essence, the Strawberry o1 model transforms real-time computing from a landscape stuffed with unpredictability and complexity into one where reliability, efficiency, and scalability is just normal.?This shift enables the development some very capable real-time scenarios across a wide range of industries and applications.

Here are just a few ideas that are completely plausible and very soon:

• Customer Service: Implementing AI that will handle customer inquiries with human-like understanding and empathy. Absolutely no need for a human.
• Ordering food: Simply calling the restaurant and having an AI process your order without misunderstandings. Or if you used an App the AI could give you a call if there was an issue with order and help you fix it.
• Cancelling Subscriptions: No more dreaded hold music or transfers; an AI could swiftly handle your request to cancel a mobile plan.
• Flight Complaints: Venting frustrations about delayed flights to an AI that can and will be able to provide solutions.

The applications are vast, and while tokens for voice are pricier than for text (about $18 an hour) the value is undeniable. Paying that amount for an hour of seamless, intelligent conversation is a game-changer. I thought about how quickly and affordably this technology has unfolded, and it left me excited that I am getting my JARVIS soon.

3. Introduction of the Complete Strawberry o1 Platform

You would think the first two announcements would be enough to rock the tech world but the moment that truly showcased OpenAI's ingenuity: a live demo of the o1 platform. Unlike the o1-preview we are all familiar with, this was a significant leap forward. On stage, they started with a drone that had nothing, no programming, no setup. In real-time, they got Chat to:

• Set Up the Server: Establishing the foundational infrastructure needed for operation.
• Programmed the Drone: Writing code from scratch to control its movements.
• Streamed Live Video: Enabling the drone to transmit video in real-time.
• Implemented a Graphical User Interface: Creating an accessible way for users to interact with the drone.

If that's all it does, you might think, 'Whatever’. But then you don’t understand where we’ve leaped from.

The fact that they could do all of this live, demonstrated not just the power of the platform but its user-friendliness and potential for rapid development.

OpenAI made a bold comparison: the jump from the o1-preview to o1 is like moving from GPT-2 to GPT-4.

I recalled my early days experimenting with GPT-2 and how amazed I was then. The prospect this leap is thrilling! If you haven't used GPT-2 or GPT-4, just know that the difference is monumental, it's like comparing a bicycle to a jet plane.

4. Venture Capitalists Are Scrambling

The final highlight stirred a mix of excitement and concern after DevDay. OpenAI's advancements mean that the model is not only getting better but also easier to use. Everything they announces allows for the rapid creation of smarter models tailored to specific applications, all within the OpenAI ecosystem.

I could hear the hearts of some startup founders and venture capitalists drop. For startups operating in the middleware layer, those focused on tuning large models, this was a seismic shift. Their unique value propositions were at risk of becoming obsolete. I thought about companies whose entire business model revolves around fine-tuning AI models and realised that many might not survive this disruption.

My first insight - Agent Workflows Are Coming

Combining the API for fine-tuning models, real-time voice-to-voice capabilities, and higher token limits sets the stage for agent-based workflows. This means we'll soon have real AI agents performing tasks autonomously on our computers. The AI agent marketing you have seen, before DevDay, from the likes of Salesforce, ServiceNow and Google is not about real AI agents. AI chat botting with a lot of effort on the use case massaging etc. However, this agent, this is real…

Imagine setting up an AI agent to work overnight, accomplishing tasks ranging from data analysis to scheduling meetings, all without human intervention. The thought is both exciting and a bit eerie. We're going to have to adapt quickly to this new reality, and I suspect the transition will feel surreal. I wouldn't be surprised to see these developments materialize by the first quarter of 2025. It's coming sooner than we think.

?

My second insight - Rapid Improvement in Model Intelligence

The accelerated pace at which AI model intelligence is improving. Each release brings us closer to what could be considered "human-level intelligence," at least in specific domains. I anticipate debates will continue about the nuances of comparing AI intelligence to human cognition, but the trajectory is clear.

I does make me ask myself deep questions: What are we doing in the face of such rapid AI advancement? Why are we here?

People will always need people. We'll still require human creativity, work ethic, and the uniquely human touch that can't be replicated—at least not yet.

While AI agents might assist us, perhaps even a fleet of them, the core economy won't grind to a halt. Lots of industries might implode or significantly change (call centres) Goods and services that matter will still need to be created by people for people.

I don't subscribe to a dystopian, Matrix-like vision of the future, and interestingly, neither does Sam Altman, the CEO of OpenAI. Sure, he has vested interests, but his perspective resonates with me. And sadly, I don’t have the equity his has.

Despite us having incredibly smart decision-makers like o1, it's not rational scream to the trees “the robots are coming!”.

There are significant physical and practical limitations for AI to collect all the information needed to make every decision. Human intuition, ethics, and creativity are not so easily replicated. That kind of general intelligence (AGI or G-AI) is not on the horizon yet.

That said, I am worried about this, about how unprepared we all might be for the changes ahead.

The speed at which AI is advancing means that our world could look very different in just a few years. Many people might not be ready for how this technology will integrate into daily life and the workforce.

?

Closing Thoughts

The OpenAI Dev Days provided more than just updates; they offer a glimpse into a future that's rapidly approaching. It's both thrilling and daunting to consider the possibilities and challenges that lie ahead.

This is my public service announcement: Pay close attention to developments like these DevDays. They're not just tech news; they're indicators of shifts that will impact us all. \

Whether you're a developer, a business owner, or just someone interested in the future, staying informed is the first step in navigating the changes to come.

This is significant kids, and I don’t think we quite get it yet. I know I don’t. We all need to start thinking and leaning into this future now. Otherwise, the change might tilt some of us a bit.

The next quarter is going to be interesting when the o1 Platform gets out to us all. Watch out for more insights from me. Get in the comments!