Autonomous negotiations

For the agents, by the agents?

First of all the concept of autonomous agent might not be clear to all reads. Autonomous agents in this case are software agents which intelligently act on behalf of another. To illustrate we can assume Alice has her network of autonomous software agents called "Alicebots" and she sets her Alicebots on their mission for her. These Alicebots are "task AI", in other words before people even dare bring up the topic of AGI, these bots do not have to have AI of human level, or even the level of mice, but rather insect level.

Alice and her personal AI swarm

So we have established that Alice has task AI called Alicebots which she manually assigns tasks to such as "buy food if the refrigerator signals a low food supply.". In programmer terms this code would be a really simple script based on if... then.. and instructing an Alicebot for Alice can be as simple as telling her personal swarm the if... thens... until very basic understanding is had by the swarm. Suppose the Alicebot in the swarm named Tabbybot receives the request from Alice to accept the conditional job of shopping for food whenever it receives the signal?

Tabbybot shops for Alice

Now we can assume that after 72 hours give or take the signal from the smart fridge signals that there are no more eggs. This signal traverses the Internet of Things home network to reach the main computer which sends the signal to the Internet and to her personal swarm. Once the personal swarm receives the signal assigned specifically for Tabbybot the bot knows that signal means it is time to activate a shopping task from it's list of saved or preset tasks. Tabbybot then would use various machine learning and AI to achieve the goal but in this case because it's only eggs, Tabbybot simply connects to a blockchain, offers a reward in the form of digital currency for any but in any swarm which can locate a seller of eggs, and once a seller of eggs is found the bots which referred the seller to Tabbybot get their commission from the reward and a direct connection between the seller and Tabbybot is made.

Now assume the seller is a company run by humans such as a dairy farm which now receives the order from Tabbybot on behalf of Alice. The company may have it's own corporate swarm of bots and now the question is what happens next?

Autonomous negotiation between Tabbybot and Okasha Farms

Tabbybot in the most simple form has the intellect of a bee. Tabbybot is acting as a scout in the swarm which has a hive memory. The memory of the swarm may be stored centralized on Alice's computer or it may be decentralized completely stored in a secure homomorphic encryption enabled DHT. This swarm in theory should also be capable of symbolic AI in that it has a shared knowledge base not just "memory", and can reason over that knowledge base to make some common sense decisions. Suppose in this shared knowledge base Tabbybot has access to legal concepts, legal language, contract theory, contract law, and more, so that Tabbybot can open a negotiation and establish a contract on behalf of Alice with any knowledge compatible bot anywhere on the Internet?

Tabbybot theoretically speaking can negotiate a price to pay in digital currency to the OkashaOrderbot. They can in fact negotiate a price better than human beings because the negotiation will be entirely rational with no influence of marketing or other tricks. The rational conversation between Tabbybot and OkashaOrderbot would allow for Alice and OkashaFarms to get the "best fair" deal. At the same time this solves the attention scarcity problem and let us assume for sake of rhetoric that Alice has attention deficit disorder and is a comupulsive shopper? The Tabbybot would save Alice time and money while also doing the same for OkashaFarms allowing for a potentially more efficient market.

Of course that is speculation and it is up to economists to look at benchmarks of any real world implementation but in theory autonomous agents due to resolving the limits of human attention, bounded rationality, and speed, can change how people buy and sell. Autonomous negotiations is a new paradigm for trade between entities (human&human, human&machine, machine&machine).

A discussion of the theory

So let's make a list below of some of the technical mechanisms which might make the above illustration possible?

• Smart fridge (probably with voice/video recognition for programming and sensors to detect food status)
• Internet of Things (probably with a DAG/directed acyclic graph or similar network)
• Digital currencies, smart contract platform
• Personal private cloud (homomorphic encryption + DHT, connected to private/public blockchains)
• Personal swarm platform (personal knowledge base stored locally and or global shared knowledge base, programmable blank slate personal autonomous agents, symbolic AI with ability to access neural network AI for a fee using markets for data, computation, and for AI).

The tools to create Alice's personal swarm exist today. Personal assistants exist now and we have gadgets which people talk to and which access a neural network on the Internet to learn. We have the ability to create a virtually unlimited amount of autonomous agents today. We have big data today and data markets are in the beginning of becoming practical. AI markets are also on the way and will be here within months. Autonomous agents will have the biggest impact on trade as we know it, and will automate negotiations. These negotiations could be legal, they could be even something like Alice is looking for a husband, it can literally be automated without the need for dating sites or third parties. It also could spell the end of shopping sites as we know it, as commerce can take place automagically.

Data mining and data markets also change the game. For autonomous agents or any AI to be useful it must be subscribed or have access to the highest quality data feeds. These data feeds represent signals. In the case of Alice we had the smart fridge which provided a signal for Tabbybot but if we recall that Alice has a personal swarm then she could have hundreds, thousands, millions of bots, all scattered throughout cyberspace, collecting data, collecting knowledge for her knowledge base, selling data, all which means markets. For example Databot could do all data mining for Alice to fuel the growth of her personal swarm, and of course indirectly fuel the personal growth of Alice. Databot could simply monitor Twitter feeds, social networking apps, to capture public sentiment signals, monitor thousands of websites for statistical data, monitor hundreds of blockchains, etc.

If we think of CoinMarketCap as nothing more than a datafeed then Databot would have the task of collecting all signals from that feed which Alice has deemed relevant.

References

Yaqub, E., Yahyapour, R., Wieder, P., Kotsokalis, C., Lu, K., & Jehangiri, A. I. (2014, June). Optimal negotiation of service level agreements for cloud-based services through autonomous agents. In Services Computing (SCC), 2014 IEEE International Conference on (pp. 59-66). IEEE.

Lin, R., Kraus, S., Baarslag, T., Tykhonov, D., Hindriks, K., & Jonker, C. M. (2014). Genius: An integrated environment for supporting the design of generic automated negotiators. Computational Intelligence, 30(1), 48-70.

Vahidov, R., Kersten, G., & Saade, R. (2014). An experimental study of software agent negotiations with humans. Decision Support Systems, 66, 135-145.