Introduction
In my quest to navigate the technological chaos that surrounds us, I've developed a mental model that I believe offers a clearer perspective: the concept of generative technology.
Generative technology offers a compelling analogy; it provides a broader perspective on the potential directions of technological evolution. We are looking at a future where tools, ranging from transformer language models to warehouse robots that operate around the clock, will autonomously perform tasks in the real world on our behalf.
This concept offers a powerful framework for imagining the future of data-driven innovations that autonomously engage with the real world. However, a significant obstacle hinders my full acceptance of generative technology, as we conventionally define Generative AI (GenAI) as the framework encompassing discussions of this kind. In this blogpost, I intend to clarify the confusion between GenAI and Generative Technology (GenTech). A subtle yet important distinction exists between them, meriting our careful consideration. As we advance into a future dominated by intelligent agents making decisions for us, understanding the differences and implications of these frameworks becomes essential.
Why merge GenAI and robotics?
I believe it's a compelling approach to collectively navigate the risks associated with Generative AI and the domain of robotics. As both domains are capable of "generating actions in the real world" and are becoming increasingly accessible to the average consumer, the need for a comprehensive conceptual framework becomes evident.
Outsourcing Intelligence: Where are we heading with GenTech?
I guess my overall take on this post is that we have to think much more broadly here. At its core, the evolution of generative technology is about outsourcing aspects of intelligence—learning, reasoning, decision-making—to non-human agents. This paradigm shift challenges us to rethink our relationship with technology, as these agents increasingly perform tasks that were once the sole domain of humans. I feel that there are many breakthrough developments that points to the need of fleshing out a kind of framework that generative technology brings, some of them have happened in the last three months!
Developed by the Rabbit AI’s Research Team, LAM (or Large Action Model) is a sophisticated system that integrates neuro-symbolic programming with state-of-the-art technologies. This model directly models the structure of various applications and the actions performed on them, offering a novel approach to AI interactions.
"Genie: Generative Interactive Environments" is a groundbreaking AI model introduced by Jake Bruce and others, capable of generating a variety of action-controllable virtual worlds from unlabelled Internet videos. This 11B parameter foundation model includes a spatiotemporal video tokenizer, an autoregressive dynamics model, and a latent action model, allowing for interaction within these environments without ground-truth action labels. Genie opens new avenues for training agents to imitate behaviors from unseen videos, aiming to develop generalist agents for future applications.
Sora's text-to-video technology is an evolution of the techniques used in OpenAI's DALL-E 3, utilizing a denoising latent diffusion model with a transformer as the denoiser. The model generates videos by denoising 3D "patches" in latent space, then converting them to standard video format. Despite its groundbreaking achievements, OpenAI acknowledges Sora's limitations, such as challenges in simulating complex physics, understanding causality, and differentiating left from right.
The first prototype of Optimus was showcased in 2021, demonstrating basic capabilities like walking and waving. However, it was the unveiling of Optimus Gen 2 in December 2023 that marked a significant leap forward. This new generation robot is lighter, faster, and more capable, featuring Tesla-designed actuators and sensors. It's designed to move 30% faster, with a weight reduction of 10 kg, and improved balance and mobility. The hands of Optimus Gen 2 have been significantly upgraded to be more dexterous, capable of handling delicate objects, which underscores Tesla's ambition to deploy these robots in varied, potentially complex tasks.
Keeping pace with the rapid advancements mentioned above can be overwhelming, as the pace only seems to accelerate. The world is increasingly recognizing that GenTech is here to stay, a fact underscored by the recent influx of funding into AI startups and Nvidia's stock price surge, which has surpassed that of Tesla's.
Hence, positioning GenTech as a conceptual cornerstone for any tool that generates a product or service, whether it be an application or a robot, offers a strategic approach to encompassing the generative capacities of these technologies. By doing so, we not only align with the essential framing of technology as an enabler but also embrace a vision that accommodates the complexities and distinctiveness of generative technologies. This enhanced perspective fosters a more coherent and forward-looking approach to the technological ecosystem, emphasizing the synergies between AI and robotics, and paving the way for innovative solutions that transcend traditional boundaries.
GenTech and Risk Management
In navigating the field of risk management, I find it more useful to approach the unknown unknowns or known unknowns through the lens of generative technology rather than solely GenAI. This perspective is crucial for risk managers to consider, especially as AI systems begin to take on physical forms. The implications of robots autonomously generating results for us are significant and warrant thorough discussion.
Shifting our perspective from GenAI to a broader conception of generative technology, which encompasses robotics, can streamline our efforts and save time. This redefinition brings clarity to previously misunderstood elements, underlining the fact that our focus should extend beyond AI alone. Generative tools, notably including service-producing robots, are set to become essential elements of our technological environment within this decade. Recognizing these advancements is crucial. Acknowledging that these tools will autonomously generate outcomes underscores the importance of adopting a holistic approach to risk management that addresses the entire range of impacts posed by generative technology.
Conclusion
This discussion serves, in part, as an exploration into the tumultuous landscape of technological advancements, particularly in how we might navigate the uncertainties presented by generative technologies in the years ahead. The prospect of living in a reality where automated systems make decisions on our behalf may have once seemed like a scenario straight out of a science fiction novel.
In conclusion, the challenge of ensuring safety within generative technology remains unresolved, prompting me to revisit my original framework of understanding. While we are yet to establish a world where generative tech is universally acknowledged as safe, this dilemma is intrinsically linked to my ongoing research into ethical alignment. Despite the persistence of alignment issues, generative technology undeniably occupies a critical space in our dialogues. Whether it will transition into mainstream discourse remains to be seen, but its potential impact is undeniably significant.
If you found this post helpful, consider clicking the share button or explore my posts on mental modeling in the age of generative AI. Additionally, if you are interested in viewing the services I offer, please feel free to browse them here. Also, consider checking out Doc Ligot’s webinars on Generative AI; the link is available here.
What is Generative Technology?