The Future of AI/Digital health: Rapid Advancements and the Critical Need for AI Expertise

As we approach the final quarter of the year, the trajectory of artificial intelligence (AI), particularly large language models (LLMs) like ChatGPT, is set to undergo profound transformations. While the advancements we’ve seen so far in 2024 have been remarkable, real game-changing developments are expected to emerge in the coming months, especially after the U.S. elections.

Several key factors will drive this progress:

1. Order of Magnitude in AI Development

The current pace of AI advancement is accelerating at a rate that can be best described as an order of magnitude – roughly a tenfold increase each year. To put this into perspective, just one year ago, GPT models weren’t capable of performing specialised tasks as they do now, and AI systems were significantly slower. Today, we see models like Gemini making strides in generating high-quality images through diffusion techniques. This rapid acceleration is expected to continue, with capabilities increasing exponentially.

2. The Impact of New Chip Technology

One of the most exciting developments in AI hardware is the introduction of the B200 Blackwell chips, which are 30 times more powerful than the H100 chips currently used to train the most advanced models. This means that the models we consider cutting-edge today could soon be outpaced by those running on significantly more powerful hardware. With the combination of faster hardware and the ongoing tenfold annual increase in AI capabilities, the next generation of models, set to be released in early 2024, will be far more advanced than anything we’ve seen.

3. Data: The Next Frontier

While much of the easily accessible data on the internet has been utilised for AI training, the future lies in unstructured, untapped, and multi-modal data sources. This includes video, sound, and image data from around the world, as well as synthetic data—artificially generated information that can be used to train other models. This data has the potential to open new horizons in AI training and development, though its full utility is still under debate.

4. OpenAI’s Plan for a Comeback

OpenAI has faced increasing competition from models like Claude, Gemini, and even the open-source LLama 3.1. However, it is preparing to regain its lead with the introduction of new models that incorporate System 2 thinking. Unlike current models, which simply predict the next token, System 2 models use a feedback loop to verify the accuracy of their predictions before presenting them. This will represent a quantum leap in reasoning and accuracy, with far fewer instances of hallucinations.

However, these improvements come at a cost: System 2 models are slower in generating responses. While this may make them less suitable for traditional end-user applications like ChatGPT, their enhanced reasoning capabilities make them ideal for Autonomous Agents—AI systems that can perform tasks independently. These agents have been held back by the limitations of current models, but the enhanced reasoning of System 2 models could unlock new possibilities.

5. Looking Ahead: The Orion Model

One of the most anticipated developments in AI is the upcoming Orion model. This model is currently being trained on synthetic data generated by a precursor system known as Strawberry. Orion is expected to represent a new level of capability, pushing the boundaries of what AI can achieve.

The Urgent Need for AI Expertise

As these advancements continue, there is an urgent need for professionals to acquire AI expertise, particularly those who are currently in their mid-20s or older. The pace of AI evolution means that many professionals will soon find themselves in senior positions, and without strategic knowledge of AI, they risk being left behind.

We are already seeing this shift in industries such as marketing, where younger professionals equipped with advanced AI tools are outperforming their more experienced counterparts. This trend will only accelerate in the coming years. Those in their late twenties or thirties who lack AI proficiency may find themselves with diminishing opportunities as AI becomes more integrated into the fabric of every industry.

In conclusion, The future of AI is evolving rapidly, and the next 12 months are poised to bring developments far beyond what we’ve seen so far. AI agents will become more capable, autonomous systems will become more reliable, and the demand for AI expertise will continue to rise.

For those looking to stay competitive in this fast-changing landscape, now is the time to invest in AI knowledge and skills. The future belongs to those who are prepared to harness the power of AI and adapt to the coming changes.