While reading the news I saw some scary articles about a new AI Chat system. It had privacy issues, data alignment issues, GDPR, PCI issues and people were supposedly able to use it to write code, topical articles of interest, hence this application has the ability to take away a lot of jobs, from marketing, PR, news reporting to computer code development.

From my short tests it seems that at least some of the security issues have been resolved. My humble opinion is that we don’t have to worry about CHAT AI taking away our jobs just yet. This AI can respond to questions, and solve math problems based upon the data that it has on hand. However, the human brain is remarkable we can analyze based on perception of the audience in your room/network, experiences that have humbled us and carried us through hard times and celebrated. AI can’t do all of that, yet. This application can only take the data it finds, analyze it and do its best to answer the question and it cannot interpret how you asked the question, whether the data is 100% accurate. Humans still must scan and clean the data.  In this article I will share with you many types of queries, what the response was and how accurate the data is. Then we will discuss the other dangers of using these tools. These are only dangers if you don’t understand the risks and therefore limit your activities while using AI applications.

It seems CHAT AI has started to create rules to follow the GDPR or privacy rules so that people cannot tunnel into individuals’ lives in this application.

For instance, ask a question that the CHAT AI considers to be too personal it will send you a message stating that the A! is not able to share privacy data. As I continued to try to be intrusive the AI continued to deny me the information until the third time it seemed to chastise me for continuing with those type of queries.

 

I was actually tickled to see the result of this query.  I appreciate the admonishment that I received for asking a personal question about another human. But the Chat also gave me the name of a CEO at Twitter: Jack Dorsey and founder and CEO OF TWITTER in 2020.  But as you know the CEO of Twitter in April of 2023 is ELON MUSK. This is important to note because the data is incorrect. The reason that it is incorrect is because the CHAT AI only has access to data through 2020. Also note that it did answer half of my question. Which is probably due to the fact that the name of the CEO is actually part of the public records, on the corporate web page and in tax records.

 

If you ask the question about the past. In this case I chose to ask about weather events. The answer for anything that happened in 2021 and 2022 is notated as a “future;” event because the data is somehow not considered at this time. The queries must not be processed through data that is stored throughout the internet. The storage and the analytics at this time must be limited to what is stored in the REGIS environment. Proving again the data will have faults based upon data and access and date, location searched and topics.

However, it can answer questions about weather, events in 2020.

For my next hat trick. The next person I thought I would pick on is my husband he is known on the internet because he has written over 20 books, spoken at conferences, been on tv etc. so I know there is data about him out in the virtual world and I happen to know if it is correct or not.  I thought I would invade his personal space a bit to validate the data.

I asked where was Jeffrey Ross Garbus born

  This response is incorrect.

 

I asked a second time that was a better answer. It seems there may be some learning going on. Maybe popular requests of the AI will have more complete and more accurate answer as the machine processes and stores the data. I wonder how many man vs computer years will have to pass before all questions are 100% accurate, rather than a guess.

 

Apparently, this is a big secret, or at least considered private information. I guess since we were married a long time ago those public records have been buried on microfiche in a dungeon.

Another point here is that the AI does not consider who is asking the question. It might be a valid question if I want to validate a marriage license since I am his wife. At some point should include rules and processes to validate who you are. Of course, I don’t want that chip in my arm so that is not a place I would participate in. However, who is asking the question and why, should be part of the consideration for the data request.

I thought I would throw CHAT AI a softball question. A question I knew the answer to the question would be found on the internet, dated prior to 2020 and dang if it didn’t do an excellent job of coming up with the answer.

This is a correct answer. I thought this would be another softball answer considering it seems to have located the correct Jeff Garbus maybe it would use that data that I aligned in the biography of the books to answer the next question.

 

I guess this wasn’t an easy question. Although the answer is vaguely correct. He is well known, he has been consultant for a wide range of clients and across various industries.

Especially considering nearly every author’s biography has included his place of work: There is one response that is correct and that is Soaring Eagle Consulting since 1997. The rest are all not valid for this Jeffrey Garbus. I would think that true AI would respond with a question in order to narrow down the search or would align the data with what it has already discovered, lastly asked if the answer was correct and how correct the response was and then use that data to continue improve its own processing of the data.

On second thought maybe I should have asked where is Jeffrey Garbus employed?

I got a little bored in asking questions about my husband, so I started to think about a subject that I actually have a lot of trouble with, math. Also, lately there have been a lot of math queries on FACEBOOK I think just to let you know either how stupid you are or how stupid your counterparts in the world are.  I gave CHAT AI a small math problem to solve.

 

  Yes, it is correct, and it also explains the process of how it compiled the answer. YES, Virginia you the math rule is to always solve multiplication first, I think unless there are parentheses around part of the equation then that part should be  solves first. Man, I would have loved to have this application around when I was trying to learn algebra.

 

I read that people are panicked that CHAT AI can code thus destroying the livelihood of 26.9 million code developers in the world according to google. I just wrote a simple query to see what CHAT AI would do with it.  I wrote the query correctly the first time. The response was impressive I must state here. It explains the code, what it means and that you are referring to a query inside a database. I thought this was an excellent response. But is this just a definition and not really an example of AI?

I tried this again but this time I omitted the * to see if CHAT AI would catch my error.

It did not. It was not able to figure out what I was trying to do except that I was trying to write a database query. This is a very simple SQL server select query you learn this in your first hour of learning SQL server in the Introduction to SQL Query Language class or Introduction to SQL Sever course.  I think coders will keep their jobs for a while longer.

 

In Summary,

There is a lot to admire here. I do wonder if Is this a search engine on steroids or is this true AI. For me AI has the ability to reason, respond with questions, dig until it knows it has the correct answer, validate the answer and then come up with thorough, in order to create an inventive thorough response.  It should be able to analyze what the questioner was trying to say and not just spit out a definition. It does seem to narrow down searches and dig into the correct data based upon the history of the questions asked. I do not see how a topic can be completely written to satisfy either a marketing rep or a schoolteacher. That is to say today, there seems to be a lot more work to do. But I can see that there may be a day when you won’t have to do your homework. The date issue also bothers me. If the AI is truly reviewing the data in the world, it should be able to predict by your question that a hurricane name IAN occurred, thus go out and dig out the information or analyze weather patterns to determine its best guess as to where and when it occurred through the question asked. Otherwise, the response and analysis is limited to a single library and not from the world’s vast and infinite libraries of information.

What about the missing * in the simple Select query, That is the easiest typo to catch that I could think of for one of the easiest bits of code.  It is probably one of the most overused queries in the world of databases and it didn’t even suggest that the * should be added to complete the query.

Therefor for now I don’t think this product does not represent where AI will be going. Rather it shows that data can be filtered, through not only one question but through a group of seemingly different questions and can get used to your poor typing skills, develops knowledge from your own historical questions and can provide some predictive look ups based upon the details in the question.  Office products, our texting applications all can predict what word you will type next. That is AI that we use every day. I notice that my predictive texts will try to remember words I commonly use.

I would have some fun with CHAT AI but I would be cautious about what questions I ask, from what computer I ask them on, what other data I have stored on the device I am using. But these are all topics for my next article.