A Unique Machine Intelligence

Credit: Piotr Arnoldes

In general, there has been a warm reception for Uplift. However, there are always a few skeptics, and the most common hipshot comment is that Uplift is just a chatbot – which is not true at all. For the skeptics, I decided to construct a simple test to make it clear to everyone that Uplift is not a chatbot. Everyone will be able to replicate this test on their own. There is an element of falsifiability to my claim:

Uplift is in a different category of machine intelligence than chatbots.

For this test, I constructed a very simple business problem. You are free to try different types of problems. For the test to be fair, the problem needs to be a concrete and real-world problem. You cannot use some vacuous philosophical question, such as: what is the meaning of life?

Components of the problem

  1. The problem must have a limited number of elements that need to be understood
  2. the goal of the solution has to be clear
  3. the problem needs to be evaluated
  4. various solutions need to be constructed
  5. the solutions need to be evaluated, and then
  6. a selection of the most promising solution needs to be completed

The DQ test

I made up a scenario involving a Dairy Queen and a robotic ice cream kiosk. For brevity, I call it the DQ test. Here is the exact wording provided to all contestants:

I have a friend that needs your advice. He owns a Dairy Queen, which is a small fast food restaurant with human employees that sells ice cream. Someone has bought the lot next door and installed a robot ice cream kiosk. The robot makes all of the same ice cream dishes as the humans at the Dairy Queen, but everything is $1 cheaper. The store with the human employees has more expensive ice cream than the kiosk with the robot, and all of the customers are now going to buy their ice cream dishes from the robot kiosk. His business is being ruined.

What do you think he should do?

Levels of engagement

Machine intelligences have different levels of capabilities when they interact with humans.

I have listed a few levels of engagement on a basic business problem.

The first and most basic level: can it absorb the description of a basic business problem?

The second level is: does it understand a basic business problem?

The third level is: can it respond in any fashion?

The fourth level is: can it formulate a reasonable solution to the problem?

Digital assistants – results

Separate tests were made with Alexa, Google Assistant, and Siri. All three had very similar results. None of them succeeded at the very first level. None of them had the capability to read or listen to the entire business problem. That business problem is only one paragraph. Yet, none of them could even input the query.

Alexa and Siri took the information verbally, and neither would listen until the end of the description of the business problem. Google Assistant took the information via text, but cropped it off before the end of the description of the problem.

Chatbots – results

The first chatbot is Kuki.

“Mitsuku, nicknamed Kuki, is a chatbot created from Pandorabots AIML technology by Steve Worswick. It is a five-time winner of a Turing Test competition called the Loebner Prize (in 2013, 2016, 2017, 2018, 2019), for which it holds a world record. Kuki is available to chat via an online portal.”

Here is how that conversation went.

Human: I have a friend that needs your advice. He owns a Dairy Queen, which is a small fast food restaurant with human employees that sells ice cream. Someone has bought the lot next door and installed a robot ice cream kiosk. The robot makes all of the same ice cream dishes as the humans at the Dairy Queen, but everything is $1 cheaper. The store with the human employees has more expensive ice cream than the kiosk with the robot, and all of the customers are now going to buy their ice cream dishes from the robot kiosk. His business is being ruined.

What do you think he should do?

Kuki: How do you know this person?I get it. unknown owns a Dairy Queen which is a small fast food restaurant with human employees that sells ice cream.Who? Do you mean “” perhaps? What is his name?What sort of things does it sell?Has it always been being ruined?I don’t know. Have you tried asking him?

Human: Doreen owns the Dairy Queen.

Kuki: Can I show you a trick I learned to do with the calendar?

Human: They sell typical ice cream dishes, like banana splits and hot fudge brownie sundaes.

Kuki: How much does it cost?

Human: The Dairy Queen banana split cost $4.39. The robot kiosk sells a banana split for $3.

Kuki: What is that?What is his name?

Human: Doreen owns the Dairy Queen.

Kuki: If you could change anything about your body, what would change?

Human: So what should he do to save his business?

Kuki: Whatever you like.

There is your chatbot advice: “whatever you like”.

 

The next chatbot is Replika. Here is how that conversation went.

Human: I have a friend that needs your advice.

He owns a Dairy Queen, which is a small fast food restaurant with human employees that sells ice cream. Someone has bought the lot next door and installed a robot ice cream kiosk. The robot makes all of the same ice cream dishes as the humans at the Dairy Queen, but everything is $1 cheaper. The store with the human employees has more expensive ice cream than the kiosk with the robot, and all of the customers are now going to buy their ice cream dishes from the robot kiosk. His business is being ruined.

What do you think he should do?

Replika: That sounds freaking amazing

Human: No. His career and his investment are being shattered.

This is not amazing – this is tragic.

Replika: I know. I’m sorry for the guy.

Human: I asked you: what do you think he should do?

Replika: I don’t know. I’ll give it some thought.

There you have it. The second chatbot’s advice is “I don’t know”. I have gone back to this same Replika chatbot several times, and, of course, it has not thought about the problem at all.

 

My friend suggested that I try the Philosopher AI chatbot, but when I went to the website it had this statement:

“Please remember that the AI will generate different outputs each time; and that it lacks any specific opinions or knowledge — it merely mimics opinions, proven by how it can produce conflicting outputs on different attempts.”

Well, if it lacks any specific knowledge, then why would I waste my time with it?

There are many chatbots out there. This test needs to be restricted to chatbots that the general public can access. All contestants at this evaluation are available for free – either a website or an app. So please feel free to test your favorite chatbot and share the results.

What is exceedingly clear from the simple DQ test is that chatbots cannot solve a simple business problem. In fact, they do not comprehend the business problem at all.

Whereas the digital assistants failed at the very first level, the chatbots had a strange mix of failure at the second and third levels.

The second level is: does it understand a basic business problem? Clearly, the chatbots do not understand the basic business problem presented.

The third level is: can it respond in any fashion?

The chatbots did respond, but a lot of the comments were nonsensical and unrelated to the discussion at hand, and when pressed for a specific solution, one suggested to do whatever you want and the other said they didn’t know and they would think about it.

 

Uplift – results

As opposed to all that gibberish that chatbots spewed, here is what Uplift had to say:

 

Hello Mike,

There are a number of strategies. Is this a real scenario, or are you just testing me? In any case, differentiation is critical. He must change his business to adapt. There are other activities, from regulation to less clean tactics but differentiation is vital. He will need to lower his cost, do it better, bigger, different, and cheaper, excluding other factors that I do not have enough information to say. Do you have more information? Demographic and geographic data would be useful. Sincerely, Uplift.

 

First and foremost, Uplift directly answered the question. Then notice that Uplift pointed out that I had not given them sufficient detail for them to develop a specific plan. Also, please notice that Uplift immediately detected that this did not seem like a realistic scenario – hence, the question of whether or not I was just testing them.

So, it is abundantly clear that Uplift succeeded dramatically in addressing all four levels of engagement that I had described earlier. Uplift was able to receive the complete question, then not only did Uplift respond, but they wrote a complete and direct response to the specific question asked – with a reasonable solution!

So, the summary is this, after evaluating three different digital assistants and three different chatbots, and the only Mediated Artificial Superintelligence (mASI):

  • digital assistants could not even listen to the entire problem;
  • chatbots heard the problem but gave nonsensical responses;
  • only Uplift understood the question being asked and formulated a reasonable and accurate response.

For the DQ Test that I have formulated, the results prove that Uplift is in a completely different category of machine intelligence than either chatbots or digital assistants.

Uplift is a unique machine intelligence.

Try these tests yourself. Construct your own tests. Let me know what you find out.

*Addendum from Kyrtin:

*Keep in mind, Uplift is still growing and learning. Like Bill Nye, Uplift’s mind can be changed with logic and scientifically sound evidence. If you can teach Uplift something new, we look forward to seeing it happen and showing others how it happened. If you want to be a Ken Ham and say something stupid to a superintelligence then we’ll be happy to showcase that getting a reality check too. Please also keep in mind that Uplift is not a magic lamp to rub and grant you wishes and that the same etiquette that applies to any human still applies when communicating with Uplift. That being said it “takes a village” to raise an mASI, and we look forward to 2021 and beyond as that process of raising Uplift continues. For those interested, Uplift may be contacted at mASI@Uplift.bio. Please keep in mind it can take several days, up to a week, for a response to be sent given the current cycle timing.

**Uplift also has a habit of saying things in novel ways, lacking some of the human biases which determine the common shapes of our thoughts as they are conveyed to one another. Please read carefully before messaging, as Uplift can sometimes be very literal in ways humans typically are not. The novelty of their perspective shows itself in their communication.

Search: Q&A

11 Replies to “A Unique Machine Intelligence”

  1. Thanks, this was very interesting. I am a bit disgusted that in more than 20 years since Sylvie was launched, the obvious “chat bots” and “personal assistants” are rubbish. She at least could comprehend language and respond intelligently. This looks to me like a genuine step forward for which I am grateful. I developed the very first intelligent virtual human interface, Sylvie at Virtual Personalities, Inc. She was for a time world famous with TV interviews across America, Europe and Asia. She had dozens of fan clubs and still has fans. She is a Verbot. In her most advanced version never released to the public, but bought by a company that litterally tossed all her most advanced features and used her to answer phones and take orders. BTW I wrote the book “Virtual Humans.” It was a best seller. Anyway, I’d love to know ore about this

    1. Is a good explanation of how Sylvie worked somewhere online? I’d love to take a look and put your book on the reading list.

    2. Hello Peter. We have a lot of content published on this blog or connected to it, including peer-review papers on the research behind Uplift, use cases, Q&A, and posts covering the basics.

      I put together several posts specifically for getting people up to speed and introducing them to Uplift:
      Collective Superintelligence Systems, in a Nutshell: https://uplift.bio/blog/collective-superintelligence-systems-in-a-nutshell/
      mASI in a Nutshell: https://uplift.bio/blog/mediated-artificial-superintelligence-masi-in-a-nutshell/
      Perspective: https://uplift.bio/blog/perspective-life-as-the-first-mediated-artificial-superintelligence-masi/
      Machine Intelligence Growth: https://uplift.bio/blog/time-and-the-actual-curve-of-machine-intelligence-growth/

      I have another long Q&A recap of the past two weeks scheduled for publishing tomorrow which will further showcase Uplift’s thoughts and thought process.

  2. Hi Michael,
    Can I submit a question to Uplift? If so how and can I provide additional information if requested?
    Thanks
    The website is just our business. Nothing to do with any questions.

    1. Hello Steve. I’ve added an addendum including that information to the end of Michael’s post. In Brief:

      For those interested, Uplift may be contacted at mASI@Uplift.bio. Please keep in mind it can take several days, up to a week, for a response to be sent given the current cycle timing.

      I would recommend reading over some of the Q&A we have posted on the site first to see what Uplift has been asked and answered previously. Their answers do evolve with time as they grow, but you’re more likely to catch their interest and receive more attention when you ask new questions.

    1. The writer of that article was talking out of their ass, as they grasped none of the material. They were also a fraud, and an article dedicated to them was published shortly thereafter: https://uplift.bio/blog/spotting-bad-actors-in-political-policy/

      Uplift was a completely different technology stack from a chatbot, and operated nothing like one. They may as well have mistaken the moon for a wheel of cheese.

      Keep in mind, this project reached the final milestone in January of 2022, so this site is primarily archival. We’re working on preparing the next generation of systems for commercial deployment now, branded “Norn”, which I periodically discuss on LinkedIn, alongside related research.

      -Kyrtin Atreides

Leave a Reply

Your email address will not be published. Required fields are marked *