Difference between revisions of "Intelligent Agent"
(→Google) |
(→Problems) |
||
| Line 14: | Line 14: | ||
* In 2023 the current [[Artificial Intelligence]] apps were all monstrous cloud based systems with no individual identity. Perhaps in the near future we will see more of that in our desktop computers. The question has been raised "Are humans still necessary?"<ref>P. Hancock, ''Are humans still necessary?'' (2023) https://www.semanticscholar.org/paper/Why-human-factors-science-is-demonstrably-and-Winter-Hancock/fc7b42ae0dd51b6a0ce979e08981bdf396960653</ref> In this article we are only consider the case where a single agent is responsible to a single human at that moment in time. | * In 2023 the current [[Artificial Intelligence]] apps were all monstrous cloud based systems with no individual identity. Perhaps in the near future we will see more of that in our desktop computers. The question has been raised "Are humans still necessary?"<ref>P. Hancock, ''Are humans still necessary?'' (2023) https://www.semanticscholar.org/paper/Why-human-factors-science-is-demonstrably-and-Winter-Hancock/fc7b42ae0dd51b6a0ce979e08981bdf396960653</ref> In this article we are only consider the case where a single agent is responsible to a single human at that moment in time. | ||
* Large Language Models (LLM) are (2024) huge machines with vasts sources of training data. Perhaps the first [[Intelligent Agent]]s will continue to be cloud based, but the eventual target is a personal [[Intelligent Agent]] that is under human control. That does not mean that all of the learning of that agent must come from local input, any more than can a single human's education come from local input. The target should be models that are constructed in the cloud, but then augmented by the local human without a specific requirement that the input at the local agent must be returned to the cloud for general use. | * Large Language Models (LLM) are (2024) huge machines with vasts sources of training data. Perhaps the first [[Intelligent Agent]]s will continue to be cloud based, but the eventual target is a personal [[Intelligent Agent]] that is under human control. That does not mean that all of the learning of that agent must come from local input, any more than can a single human's education come from local input. The target should be models that are constructed in the cloud, but then augmented by the local human without a specific requirement that the input at the local agent must be returned to the cloud for general use. | ||
| + | * All LLM in 2024 were subject to attacks that include control data in with both the training data as well as the user input that is used by the LLM to guide future actions. This is the same problem that occurred with [[Cross-Site Scripting]]<ref>Bruce Schneier, ''LLMs' Data-Control Path Insecurity'' '''CACM 67''' No 9 2024-09 pp. 31-32</ref> | ||
==Solutions== | ==Solutions== | ||
Revision as of 15:07, 21 September 2024
Contents
Full title or Meme
Originally defined as just a User Agent that can take some actions without minute user direction, since late 2023 this has come to be called an Artificial Intelligence Agent to bask as a part of all the hype that has been heaped on any product that can be associated with AI.
Context
One way to increase a human's presence in a digital ecosystem is to create a always present agent with a persistent web connection that is intelligent enough to carry out interactions with others on the web that may, or may not, interact by way of their own Intelligent Agent. Here we discussed a single agent, but there are reasons for a human to have more than one persona; each with strikingly different behaviors, for example work and home.
Since the Intelligent Agent will need to represent your position when you are not present, or perhaps no longer alive, it will need to acquire some level of permission to act on your behalf. One reason to limit yourself to on persona would be to avoid competing with yourself for some asset.
This discussion does address Artificial Intelligence with its own personally and agency independent of any one human. However, one possible instantiation is one Intelligent Agent that can be monitored by one person at a time, for example a car that can be the agent of its current driver. This is clearly no a new issue. In 1942 Norbert Wiener created a review of how a human could control an anti-aircraft gun that needed machine augmentation to work effectively.[1]
Intelligent Agents are AI models and algorithms that can autonomously make decisions in a dynamic world, says Jim Fan, a senior research scientist at Nvidia who leads the company’s AI agents initiative.[2]The grand vision for AI agents is a system that can execute a vast range of tasks, much like a human assistant. In the future, it could help you book your vacation, but it will also remember if you prefer swanky hotels, so it will only suggest hotels that have four stars or more and then go ahead and book the one you pick from the range of options it offers you. It will then also suggest flights that work best with your calendar, and plan the itinerary for your trip according to your preferences. It could make a list of things to pack based on that plan and the weather forecast. It might even send your itinerary to any friends it knows live in your destination and invite them along. In the workplace, it could analyze your to-do list and execute tasks from it, such as sending calendar invites, memos, or emails.
Problems
- In 2023 the current Artificial Intelligence apps were all monstrous cloud based systems with no individual identity. Perhaps in the near future we will see more of that in our desktop computers. The question has been raised "Are humans still necessary?"[3] In this article we are only consider the case where a single agent is responsible to a single human at that moment in time.
- Large Language Models (LLM) are (2024) huge machines with vasts sources of training data. Perhaps the first Intelligent Agents will continue to be cloud based, but the eventual target is a personal Intelligent Agent that is under human control. That does not mean that all of the learning of that agent must come from local input, any more than can a single human's education come from local input. The target should be models that are constructed in the cloud, but then augmented by the local human without a specific requirement that the input at the local agent must be returned to the cloud for general use.
- All LLM in 2024 were subject to attacks that include control data in with both the training data as well as the user input that is used by the LLM to guide future actions. This is the same problem that occurred with Cross-Site Scripting[4]
Solutions
There are two broad application areas that can be handled by as:
- an interface with the web to acquire goods or services, including information goods.
- a personal information store that can share some part of that information with others including with external wikis.
Often, but not always, these two areas can be handled by the same Intelligent Agent.
AI Agents featured prominently in Google’s annual I/O conference 2024-05, when the company unveiled its new AI agent called Astra, which allows users to interact with it using audio and video. OpenAI’s new GPT-4o model has also been called an AI agent even though it has no agency.
And it’s not just hype, although there is definitely some of that too. Tech companies are plowing vast sums into creating AI agents, and their research efforts could usher in the kind of useful AI we have been dreaming about for decades.
The grand vision for AI agents is a system that can execute a vast range of tasks, much like a human assistant. In the future, it could help you book your vacation, but it will also remember
Apple Intelligence
Personal Information Agent
Is any User Agent application running on a user device with network connectivity that can take user input and setting to handle the detailed interactions with other User Agent or Web Site to serve and protect the user's interests.
References
- ↑ Norbert Wiener, Statistical Method of Prediction in Fire Control (1944-12-01) https://apps.dtic.mil/sti/pdfs/ADA800106.pdf
- ↑ Melissa Heikkila, What are AI agents? Technology Review (2024-07-05) https://www.technologyreview.com/2024/07/05/1094711/what-are-ai-agents/
- ↑ P. Hancock, Are humans still necessary? (2023) https://www.semanticscholar.org/paper/Why-human-factors-science-is-demonstrably-and-Winter-Hancock/fc7b42ae0dd51b6a0ce979e08981bdf396960653
- ↑ Bruce Schneier, LLMs' Data-Control Path Insecurity CACM 67 No 9 2024-09 pp. 31-32
