Does AI and Chatbots have a place in Hockey?

February 4, 2023
| No Comments

There has been lots of discussion about A.I. (Artificial Intelligence) with feeds being populated about OpenAI and chatbots as of late. Unless you’ve been living under a rock for the past few months, the global buzz over ChatGPT, the cutting-edge new helper bot designed by the Elon Musk-backed tech firm OpenAI is everywhere.

My career in Digital Marketing allows me to be familiar with a lot of technology. I’m well versed in various algorithms and analytics, and I would consider myself an earlier adopter of said technologies. However not all new technology has proven beneficial. Especially when it comes to AI. As an example, one could say that Google’s search results were much more relevant before they incorporated their AI based algorithm RankBrain in 2015. (RankBrain is a machine learning (AI) algorithm that Google uses to sort the search results and help them better understand search queries. More on RankBrain here) I’ve worked in Search for 20 years and I can say that depending on your search query, Google’s search results do not always hit the mark especially in recent years. Now we’ve also got AI that prepares online content although not without issue.

According to some, ChatGPT has become so good at its job that experts fear it could render Google and many jobs obsolete — it’s even been blocked at NYC schools because of its efficacy as a cheating tool.

So with all of the recent discussion about ChatGPT, I needed to test it our for myself. According to OpenAI, theu describe their ChatGPT as a tool to “optimize language models for dialogue”. Their “dialogue format” makes it possible for ChatGPT to answer followup questions, admit its mistakes, challenge incorrect premises, and reject inappropriate requests. (See https://openai.com/blog/chatgpt/ for additional information on OpenAI’s ChatGPT).

I asked ChatGPT a series of hockey related questions and here’s what it came back with. You can be the judge of how effective it is.

Open AI’s ChatGPT Response to Hockey Questions

Here’s a sample of question that I asked of ChatGPT. These are unedited in their form. It is worth mentioning that CHATGPT was trained as of later 2021, so some results do not factor in information that transpired in the past 14 months or so.

The first question I asked was “What are the ten most valuable Wayne Gretzky hockey cards?”

The results are not bad, but do generate more questions than answers. For example, is the assumption here that all of these hockey cards are graded and scored a perfect 10? Which company were they graded by? While ChatGPT does include a caveat that the values may fluctuate based on the condition of the card, are these values accurate? We do know that a 1979 O-Pee-Chee Wayne Gretzky rookie card sold for $3.75M in 2021. On average most of the cards listed sell for considerably less. Also if you look at the result there are duplicates of the same card but with different values. (See item #1 and #7). They are the same card but with considerably different values. So for this query I’d have to give ChatGPT a fail.

I asked “Which NHL player was drafted the latest and had the most points?” The reply was totally bizarre stating that “Joe Thornton was drafted late…” when in fact Joe Thornton was the first overall pick in 1997. (See our post of All-Time NHL Draft Picks by Team). The AI-based reply actually states “Thornton was selected first overall in the 1997 NHL draft, but that was considered a relatively late pick for a player who would go on to have such as successful career…” Uh what? Hmm AI gets it wrong again.

I also asked OpenAI ChatGPT “What is the goal song of each of the 32 NHL teams”. While the AI admitted it didn’t have a complete list it did provide some insight:

So not a poor response, just not a complete one.

I asked the bot, “Which five NHL teams are worth the most money?” The bot did pretty well with this one:

Similarly I asked “What is the value of each of the seven Canadian NHL teams?” The bot was polite enough to apologize for not having up-to-date information as its training was cut off in 2021. The bot quoted Forbes values (most likely from 2020 data).

I then asked the AI “When will a Canadian-based NHL team win the Stanley Cup?” You have to appreciate the honest answer of “I’m an AI language model and do not have the ability to predict future events…” well what are you good for then? Kidding… or am I?
Speaking of Stanley Cup question. I also asked OpenAI ChatGPT that all important “Why haven’t the Toronto Maple Leafs won the Stanley Cup since 1967? Leafs fans are you ready for the response? This was actually one of my favourite replies that I got back. The bot starts off with “The Toronto Maple Leafs have not won the Stanley Cup since 1967 due to a combination of factors…” Apparently the Leafs have had some inconsistent player performance, a number of coaching changes and injuries to key players.

There you have it Leafs fans. Any questions?

I proceeded to Ask the ChatGPT, “Who is the better hockey player? Connor McDavid or Auston Matthews? The AI couldn’t give me an awswer opting to sit on the fence with a “…both players bring unique strengths to the ice and have proven to be the best players in the league…” Uh wrong. Connor McDavid is the best player in the game period.

Connor McDavid

Speaking of Connor McDavid, I asked the Artificial Intelligence “How many points would Connor McDavid have per season if he played in the 1980s?” A bit of a vague answer here, but the bot knew enough to state that “it’s possible McDavid would still have been a dominant scorer in the 1980’s”.

I also asked who the AI thought was the greatest hockey player of all time? While the AI stated that this was a matter of personal opinion and was subjective, they did get this one right suggesting that Wayne Gretzky is the greatest player of all-time. See we knew that Gretz is still as popular as ever.

I asked “Who are the ten best female hockey players of all-time”? The bot did a fairly good job on this one recognizing Hayley Wickenheiser, Angela James and Cammi Granato as the top three.

I asked the bot “What is the ideal number of teams that should be in the NHL? It did not really provide an answer stating that it was a matter of personal opinion and depend on a variety of factors. Sheesh hopefully Bettman doesn’t use this thing for any future NHL expansion plans.

I proceed to ask the ChatGPT “Who is the most influential person in the NHL today? It’s answer included Roberto Luongo and Brian Burke. Hmm…

I then asked “Who is the most influential female in hockey today?” AI intelligence included Manon Rheaume who may have been a strong influence in the mid-to late 90’s, but not so much today. The other women listed were great choices.

I did ask which NHL team that the bot thought would draft Connor Bedard. Not a great reply on this one:

One of the better replies I received was when I asked ChatGPT “What are the top five hockey movies of all-time?”

I asked the bot questions around which franchises were the most successful in the WHL, OHL and QMJHL. The replies were pretty good for this one as well:

I had one more question for the Chatbot. “Should the Vancouver Canucks won the Stanley Cup in 1994?” To which the bot replied:

So while there is some “coolness” to this new technology not sure what the beenfit will be for finding information about the NHL or the game of hockey.

All of the technies and even some marketing folk are all over machine learning and AI and chatbots. It’s the shiny new object in the room and many are clamouring for more, but is it truly what the world needs? Do we need hyper trained Artificial Intelligence to make things more “efficient”? We have already seen this bewildering trend of tech companies having massive layoffs amidst record profits. How far are we from the tipping point of machine learning tech? Or perhaps we are at the beginning? Will Rogue AI take over and kill everyone like scientists are warning?

If interested, you can try OpenAI ChatGPT by visiting: https://openai.com/blog/chatgpt/

Fans