top of page

Autonomous AI Agents are Likely the Future. Who's Afraid of Them?

Last week, two companies once again shook up the dynamic ecosystem of generative artificial intelligence worldwide.

The first, OpenAI, simultaneously upgraded the GPT-4 model to GPT-4O and made it publicly available for free. The updated chatbot demonstrated, for the first time, almost human-like conversation capabilities in multiple languages, including recognizing the speaker's emotions, processing multimodal information (voice, video, image), all at unprecedented speed. The second, Google, introduced Gemini 1.5 Pro, which can read and process more information than ever before, including video, audio, and long texts. This model will also soon be available to everyone.

The future of artificial intelligence is very exciting

The range of expected uses for these new models exceeds imagination. For example, the video released by OpenAI shows a blind person conversing with GPT-4O on his mobile phone to hail a cab or "see" what the ducks swimming in the lake in front of him are doing...



Let's set aside the discussion on whether such company demonstrations meet reality or not. I assume that soon these two will fulfill their promise and that the ecosystem (Anthropic, Apple, Facebook, Microsoft, and others) will not lag behind for long.

5 New Uses for AI Applications After the Giants' Announcements

Now, let's imagine the possible uses for organizational training, classroom learning, and researchers and lecturers at universities. What could we do with a large language model (LLM) capable of fluent conversation in a natural language (yes, including Hebrew) within context, analyzing everything the camera captures, finding information (both trained and from the web), and maintaining a coherent conversation:

A. We could ask the chatbot to be our private language tutor. We would build a learning plan together, set goals, and desired practice methods, and start learning at our own pace. We could ask it to be a specific character, like a flight attendant at an airport in England or our grandfather who, due to dementia, can only converse in Romanian. We would practice fluent conversation, including intonation and context.

B. We could feed the chatbot all the information a new employee at the company needs, including training videos, official documents, meeting summaries, and even the code of conduct for hosting foreign donors. The chatbot would serve as a personal assistant helping the new employee quickly get up to speed and succeed. Similarly, a new teacher at a school or a new vice principal could receive close personal guidance from a virtual character that acts as an acclimation agent to the organization.

C. Teachers striving for excellence while avoiding burnout could use the chatbot to build personalized learning plans for their students, including creating custom presentations and learning videos. They could also develop bots that accompany students, providing rich and constructive feedback, and adapting to each learner. But, does this make teachers redundant? Apparently not!

An inspiring example of such a model was seen in a video by KVUE ABC news.

In a fascinating report on a private school (Alpha Schools network) in Austin, students are seen learning core subjects on computers during the first two hours of the day with an AI-based bot. Teachers use this time to guide and facilitate self-directed learning. For the remaining six hours, they practice life skills with educators, including social and emotional skills, starting from kindergarten!

If you watch this video, you'll see that the learning material includes historical figures brought to life via avatars who tell their stories. A fifth-grader compares this learning method to another school and explains that the bot explains why the answer is correct and provides immediate, reasoned feedback. Moreover, the pace and content are personalized for each student, exactly as educators have dreamed for years but were unable to achieve.


What about coverage? Are they "covering the material"?
Yes, and in a third of the time required in a "regular" school.


D. We could walk around a place and ask the chatbot to describe it as it looked in the past century, how it would look during a prolonged drought, or if the sea level rose by two centimeters (!). We could also soon ask it to create a short video or presentation describing this visually.

E. We could turn the chatbot into an agent that performs tasks related to information retrieval, editing, sending emails, summarizing meetings, booking doctor appointments, analyzing long documents, and more. All without opening our computer, just through our smartphone.

Autonomous AI Agents Are Likely the Future of the Field

But, artificial intelligence threatens jobs.

In an article in Geektime magazine, Chagai Shachar talks about the arrival of autonomous agents that will take over tedious tasks and create a new vision of work with less effort. The article raises many questions about the future of jobs that will disappear in the coming months, or at most, in the next few years. For example, it mentioned that the Bank of Israel recently published a report indicating that 34% of jobs in Israel will disappear. Following the severe report, a discussion was held in mid-May in the Israeli Knesset in a committee that dealt with the future of work with the development of AI: a key to productivity or mass unemployment? Experts in the committee were divided regarding the future of work with AI development. In my opinion, the Bank of Israel and the esteemed committee referred to the impacts of autonomous agents. This is likely the big thing that will change everything.

So, what is the difference between the LLM as we knew it until now and the future – an autonomous agent?

Autonomous agents use the ability to understand a complex task in natural language and then:

  • Use the knowledge they were trained on, combined with information from web searches to:

  • Build an action plan to accomplish the task in stages without us asking them to

  • Evaluate the execution of each stage themselves and correct it if necessary

  • Use the tools they need to complete the task without us directing them to do so


For example, if I want to plan an event for 400 people with specific preferences, budget constraints, food delivery, refrigeration, cooking, and more, I can describe the goal in detail and the agent will do the web search, review food suppliers and required unique products, cooking times, and the need for additional staff and equipment, and present everything to me with precise calculation within minutes, including alternatives and styles for my choice. I don't need to teach it to calculate, explain how to use Excel, or provide documents on food allergy risks.


The feedback loop (reflection) essentially allows the machine to correct itself, and combined with the ability to use tools, the machines mimic human action. This is quite different from using GPT as we are used to today.


So which agents already exist today? According to experts, they include: Co-Pilot, Shopify, GPTs

I watched this TED talk to learn about the difference between the familiar chatbot and autonomous agents and now I am even more curious about the possible uses. So much so that I actually intend to create one of my own.


So, what do you think about the dangers and challenges ahead?


I would be delighted if you join our website by logging in to subscribe and participate in discussions on the topics raised here.


Additionally, you can learn at your own pace in the online course available on the site, which includes 60 short recordings that will guide you step by step to understanding the field and creating your first project incorporating AI tools.


Furthermore, remember that you can book an inspiring lecture on any topic covered in the many posts on this site.


Good luck, 21st-century educators!


35 views0 comments

Comments


bottom of page