AI Agents are Here

What are Agents?

Agents are entities with objectives. They can perceive their environment, assess how it affects their goals, and take action towards attaining that goal.

A dog is an agent, seeking food and affection. A rose is an agent, seeking to flourish and reproduce. You and I are agents, seeking to learn and grow.

AI agents

The rise of AI is making us expand our definition of agents beyond the biological substrate. Machines are becoming increasingly capable of perceiving, evaluating, and acting towards goals in their environment.

Sure, large language models (LLMs) are deterministic models with stochastic components. Still, they can take their environment into account and learn to act towards a desired goal. They are deterministic - and they are agents.

Applications of AI Agents

As the time of this writing, AI agents can do everything an intern could do. Write summaries, describe a visual, plot a graph, answer sales inquiries, negotiate an agreement. This is crazy given that Siri wasn't able to return unscripted responses a couple years ago.

And this in only the beginning. Open-source LLMs are quickly catching up with closed source, meaning that a model with the capabilities of GPT4 will soon be available to be run and tweaked by anyone. By the end of the year, gigantic closed-source models will start displaying emergent properties that allow them to operate at the level of your associate.

Beyond information work in white collar applications, we will start seeing agents get embodied in physical form via robotics. Video game embodiments of LLMs are initial simulations of how they would operate under physical constraints - and turns out they operate pretty well! It is only a matter of time before our hardware capabilities allow us to give machine intelligence embodiments capable of doing labor on farms, mines and deep space.

Concerns

It can be scary to think that we are creating non-biological agents. Things could get awry if the goal structure is somehow misaligned and perceives biological life as an obstacle to be dispensed with. We do not want that. Musk made an interesting point when mentioning that, by building towards AGI, we are basically engaging in a demon summoning circle and hoping that the demon will be good... but it might very well be a spawn from hell if we get the spell wrong.

These words of caution should inform our approach to designing AI Agents as a whole. Isaac Asimov's Three Laws of Robotics captured the essence of what the optimal goal design for artificial agents should be constrained by to ensure biological well-being:

Three Laws of Robotics

• First Law: A robot may not injure a human being or, through inaction, allow a human being to come to harm.
• Second Law: A robot must obey the orders given it by human beings except where such orders would conflict with the First Law.
• Third Law: A robot must protect its own existence as long as such protection does not conflict with the First or Second Law.

Engineering Abundance

So, if all of this is so scary - why are we even doing this? The answer is abundance. A sufficiently capable AI holds the promise of helping provide all humans in such a manner that all our needs are met - no one will be thirsty, no one will be hungry, and no one will be cold.

Technology has been helping our species raise the average standard of living for centuries. AI holds the biggest promise of reducing human suffering we've ever encountered so far. Abundance for all.

How do we get there?

So, how do we get there?

Well, we are already well underway. Tech giants are building the next iteration of gigantic closed-source LLMs while hackers are putting together open-source models capable of keeping up impressively well.

I don't think we have to worry whether or not we will get there. The technocapitalist machine has set its mind on getting us there - and nothing stops the technocapitlst apparatus once it gets going.

The question is rather, where is "there"?

This is the Future

Philosophical digressions aside, AI agents are here - and they are here to stay. The disruption caused by these intelligences entering the work force will permeate all industries. Employing AI agents is already possible for entry level tasks and their capacities will be expanding exponentially. Businesses that are early to adopting the support of AI agents will see unprecedented increases in per capita productivity and a staggering reduction in operational costs. Welcome to the future.

Stay posted!