A lot is happening in the field of AI and a lot has happened in a very short time. But what can we expect in the coming years?
(I am going to ignore the assault on the data since it was already explained here in other articles).
- Adoption: First, a massive adoption thanks to cultural and social change. More people and companies will understand how it works and will take more potential from this, which will lead to increase the global productivity. Don’t forget that Manuel Jalón invented the mop in 1964, whose technology was a rag attached to a stick. Humans need time to put the pieces together and give them value.
If it took us centuries to combine a stick and a rag to create the mop, just imagine what we could achieve once we fully understand the potential of combining AI and Blockchain.
2. Open source: Open source AI models are inevitable and are here. Tech giants Meta and Microsoft have unveiled Llama 2, an open-source model for generative artificial intelligence and predictive language. Llama 2, the successor to the original Llama model introduced in February for academic and research use, is now freely available for commercial use and research. The model, which uses a sequence of words to automatically generate text in over 27 languages, has been designed to enable developers to transform the code into user applications. With over 70,000 billion parameters, Llama 2 promises improved results and more natural language. The release of the open-source model comes after the first version received over 100,000 access requests. Llama 2 is available on Microsoft’s Azure cloud services, allowing developers to leverage its native cloud tools for content filtering and security features. The code is also optimized to run locally on Windows, providing a seamless workflow for developers as they deliver AI generative experiences to customers across different platforms. In addition to Windows, Llama 2 is also accessible to developers on Amazon Web Services (AWS) and Hugging Face, among others.”
This giant leap also fits with the blockchain philosophy and makes it easier for us to stick the stick to the rag to create the mop.
3. Synthetic data: AI will not need to learn from humans. Tech giants Microsoft and OpenAI, along with other major players, are turning to synthetic data to train large language models (LLMs), as traditional web data is deemed ‘no longer good enough’ and ‘extremely expensive’, according to a Financial Times report. The move towards synthetic data is driven by the need for unique, sophisticated datasets to enhance AI performance, as the limitations of human-made data become apparent. Companies like Cohere are employing AI models to generate synthetic data, while Microsoft’s research team has demonstrated the potential of synthetic data in training smaller models. However, the effectiveness of synthetic data in enhancing the performance of larger models like GPT-4 remains to be seen. Despite potential risks, such as ‘irreversible defects’ in AI models trained on their own raw outputs, industry leaders are optimistic about the future of synthetic data. As we move into a new era of AI development, the reliance on human-created content could diminish, paving the way for a world dominated by AI-generated data and content.
From a societal perspective, this represents a significant psychological shift: AI will no longer need to learn from us, and we will no longer serve as teachers. While humans will still be required for oversight and control, even these roles could eventually be replaced. With each passing day, our necessity diminishes. Welcome to the science fiction world we once could only imagine.
4. Quantum leap: In a groundbreaking development, a Chinese quantum computer, Jiuzhang, has been reported to perform AI-related tasks at a speed 180 million times faster than conventional supercomputers. This remarkable achievement was announced by a team led by Pan Jianwei, often referred to as the ‘father of quantum’. The team suggests that the quantum computer could revolutionize various fields, including data mining, biological information processing, network analysis, and chemical modeling research. To put this into perspective, Jiuzhang completed a task in less than a second that would take the world’s fastest classical supercomputer nearly five years to solve. This leap in computational power could herald a new era in artificial intelligence and data processing, opening up unprecedented possibilities for scientific and technological advancement.
As we stand on the precipice of a new era in AI, marked by widespread adoption, open-source models, synthetic data, and quantum leaps in computational power, we are witnessing the transformation of the world as we know it. The roadmap for the next big leaps in AI is clear, and it promises a future where AI is not just a tool, but an integral part of our society and economy. As we navigate this uncharted territory, we must remember that while we may no longer be the teachers, we remain the architects, the visionaries, and the ethical compass guiding AI’s evolution. Welcome to the future, a world where the science fiction of yesterday becomes the reality of today, and humans will steel pass the mop.
Yours in crypto and AI.