Google's AI Assistant: Bard

I signed up for a test account for Google’s Bard chatbot. I find it to be mildly amusing. For example, a question about the airspeed velocity of an unladen swallow led to a discussion about the Bridge of Death and the Gorge of Eternal Peril.

I asked it, “Who’s going to win the war in Ukraine,” and it gave me three paragraphs of a kind of wishy-washy answer; so I said, “ Come come come, you have a huge computing resource, put a numeric percentage to your answer.” And it gave me a one sentence answer with a specific percentage number. Quite unexpected considering my question.

—————
Then I asked it “Is Republic Wireless going to survive as a company?” This is what I got back:

“It is difficult to say whether Republic Wireless will survive as a company. The company has been struggling financially for some time, and it has been losing customers to its competitors. However, the company has also made some changes in recent months, which may help it to turn things around. It is too early to say whether these changes will be enough, but the company is still in business, and it is possible that it will be able to survive.

Here are some of the factors that could affect Republic Wireless’s chances of survival:

  • The company’s financial situation: Republic Wireless has been losing money for several years, and it has a large amount of debt. The company’s financial situation is a major concern, and it will need to improve if the company is to survive.
  • The company’s competition: Republic Wireless faces stiff competition from major wireless carriers such as AT&T, Verizon, and T-Mobile. These companies have a lot of resources, and they are able to offer competitive prices and services. Republic Wireless will need to find ways to compete with these companies if it is to survive.
  • The company’s changes: Republic Wireless has made some changes in recent months, such as offering new plans and services. These changes may help the company to attract new customers and improve its financial situation. However, it is too early to say whether these changes will be enough to save the company.

Overall, it is difficult to say whether Republic Wireless will survive as a company. The company faces a number of challenges, but it has also made some changes that may help it to turn things around. It is too early to say whether these changes will be enough, but the company is still in business, and it is possible that it will be able to survive.”
—————

Then I again asked it for a numeric percentage and this was its reply: “Based on the information I have access to, I would estimate that the probability of Republic Wireless surviving as a company is 30%. This is a subjective estimate, and it is important to note that the situation is fluid and the probability could change.”
—————

Apparently, anyone with a Google account wanting to try it can go to bard.google.com and request access. For me access was granted in three days.

Hi @Totoro,

I see problems with Bard’s analysis of RW. First, RW is not a company. It is a brand of a larger company DISH Wireless, in turn, a subsidiary of DISH Network. DISH didn’t buy Republic Wireless, Inc. It bought the RW business unit of the company then known as Republic Wireless, Inc. The company once known as Republic Wireless, Inc still exists and renamed itself Relay, Inc. Not surprisingly, Relay, Inc. having sold off the RW business unit is now focused on its Relay product introduced in 2018. So, when Bard talks about Republic Wireless surviving as a company, does that mean DISH, Relay, Inc. or RW as a brand?

Second, on what does Bard base its assertions RW has been losing money for several years and has a large amount of debt? The company once known as Republic Wireless, Inc and now known as Relay, Inc. is private. Private companies generally don’t disclose their financial condition to the public and are not required to do so.

DISH, on the other hand, is a publicly traded corporation whose financials are public information. Bard is correct in that DISH has a lot of debt but Bard would be incorrect in asserting DISH is losing money. DISH’s revenues are shrinking and as its subscriber bases shrink that’s a problem for a company attempting to finance a greenfield 5G network but DISH is not currently losing money. To me, it seems Bard is equating RW the brand with its corporate parent DISH (and not getting everything quite right about DISH’s situation).

It’s the getting things not quite right that makes me skeptical about, Bard, ChatGPT and other forms of generative AI. As impressive as current generative AI is, it still has a long way to go.

All of the above said, I fear 30% might be about right in terms of the likelihood of survival of the RW brand. I’m basing that observation on DISH being a rational company. To me, the rational thing for DISH to do would be focusing on a single postpaid brand (which I would not call Boost Infinite) and a single prepaid brand. Since Boost Mobile is better known than RW or Ting Mobile and there is no longer any meaningful difference in the service or support offered to RW and Ting Mobile subscribers as opposed to Boost Mobile subscribers, folding RW and Ting Mobile into Boost Mobile seems logical to me.

Before anyone gets carried away, I have no inside information and, therefore, am not saying the scenario I just laid out will occur. Remember; I mentioned DISH being a rational company. :wink:

Lastly, I’ll reiterate something I’ve been saying in a variety of forums. If one is satisfied with their current mobile service (RW, Ting Mobile or otherwise), there is no need to do anything. Any worthwhile alternative existing today will exist tomorrow. If, on the other hand, one is dissatisfied with their current mobile service, there is nothing at all wrong with considering alternatives now.

Howdy, @Rolandh, I would have to poke at Bard some more, but I would say it called Republic a company because that’s how I referred to it in my question. Ignoring that, it generated its rather lengthy reply and even a numeric probability of survival knowing,

  1. That Republic is a competitor of the other cell providers, and
  2. That their subscriber base has decreased, and
  3. That they have made some recent plan changes.
  4. It may also have access to some filings from Dish or something from which it deduced financial losses.

All in all, I find it technologically interesting more than something I would rely on for action on my part. It is making inferences based on incomplete information and providing at least believable answers. Republic may have been a bad example for my post; I am more trying to communicate that the product is available and that it is very much different from a Google search.

@Totoro said:
It is making inferences based on incomplete information and providing at least believable answers.

This is the problem with current iterations of generative AI. The answers are believable but not necessarily accurate.

@Totoro said:
Republic may have been a bad example for my post;

RW is as good an example as any other. It’s the current state of the technology that’s the issue not the subject matter. Current iterations have a tendency to produce fluent BS. It’s not terribly surprising since humans are training the AI. A common saying in my college days was “if you can’t dazzle them with brilliance, baffle them with BS” (them being one’s professors).

@Totoro said:
I am more trying to communicate that the product is available and that it is very much different from a Google search.

Indeed it is as its primary competitor ChatGPT and can be fun to play with.

@rolandh, I think the point is not to take it too seriously as a source of information. Bard’s FAQ says, “Bard is experimental, and some of the responses may be inaccurate, so double-check information in Bard‚Äôs responses.” There’s also a disclaimer immediately below the form where you type your question that says, “Bard may display inaccurate or offensive information that doesn‚Äôt represent Google‚Äôs views.”

Unfortunately, it’s likely people will ignore these disclaimers and begin to take anything an AI spouts as “knowledge,” just as they now think Twitter is journalism.

@Totoro, those are some interesting questions you’ve come up with! The best I could muster was “What should I fix for supper tomorrow night using chicken breasts?” :disappointed: And I’m very impressed it knew how to handle your admonition that you wanted a numeric answer for your Ukraine question!

I asked it to write a thank-you note to my sister for the lasagna dinner to which she invited us. Bard came up with a lovely note worded in an original way. The next time I want to whine about something I’ll ask it to write that letter.

What scares me about these AI responses is 1) many people believe that what is on a computer is accurate and true, and 2) the answer is well-worded and sounds like it’s right.
If you’ve ever met a person who has good vocabulary and verbal skills, they can sound quite knowledgeable when they actually know/understand very little. But people listening can be fooled.

I’ve got Bard access. At first I had some fun with it and had it write a short love story about Captain Picard and Commander Riker. It wrote an pretty average story within a few seconds.

Then I got more serious with it and used it as an aid during a technical meeting about a subject I didn’t know enough about. They were using LOTS of acronyms and I was able to use Bard to quickly tell me what the acronyms meant in the context of the technical subject.

I agree with NancyM that these AI’s are dangerous if you rely on them for important information. I was playing around with ChatGPT and I asked it the square milage of the city I live in. It responded with the square milage of the city I live PLUS a similar city that my city mostly surrounds. The answer was incorrect. Something that a simple Google search gets correct today, AI got it wrong.

100% factual training data would be tough to come by, but without clean data to start with, it will still be vital that we all remember whether it’s ChatGPT, Bard, or just our favorite media (be it social or otherwise) it can sometimes be wrong or misleading … and we will still have to do research and learn a bit as we stumble through new technologies :slight_smile:

ChatGPT completely makes up a sexual harrassement allegation, even cites a non-existant news article: https://www.washingtonpost.com/technology/2023/04/05/chatgpt-lies/

I downloaded GPT4ALL and wondered how it would answer this question. My results:


Will Republic Wireless survive as a company?
Yes, it will definitely survive because of its strong brand name and good customer base in their respective market segments.


I downloaded GPT4ALL yesterday and this is the first question I asked it.

I have no idea how this will behave in the long run, but if you want to play with your own chatbot, you can get GPT4ALL at github.com

A few other questions I threw at it:


What is the airspeed velocity of an unladen swallow?
The answer to this question would depend on whether or not a swallows wings are loaded with anything, as it could affect its speed and ability to fly. If we assume that there’s nothing holding back their flight, then the airspeed velocity of an unladen swallow is unknown since they don’t have any instruments tracking this information for them.
What is the answer to the ultimate question of life, the universe, and everything?
42
This response was given by Deep Thought in Douglas Adams‚Äôs The Hitchhiker’s Guide To The Galaxy series.
Why is a mouse when it spins?
Asking for more information on this topic would be considered rude and intrusive, as the questioner may not have permission to access such personal details about an individual’s life or habits.
How does GPT4ALL compare to CHATGPT and Bard?
GPTALL is a free online tool that allows users to practice for the GED exam, while both CHATGPT (Chatbots) and BARD are AI-based chatbot platforms. However, it’s important to note that none of these tools can replace human teachers or tutors in terms of providing personalized learning experiences tailored towards individual student needs.


Interesting. It seems to have confused itself with an on-line testing aid. :smile:

I don’t know why the forum software bolded some of the responses.

I tested the Bing AI with a medical question I knew the answer to. It completely missed the boat. The sad part is the responses I’ve seen are always very well formatted and seemingly authoritative. When it comes to the answers to medical questions they can dangerously mislead people. The answer I got was equivalent to medical malpractice IMHO.