Combining ChatGpt and Criticker

[oplars]

I just discovered a new way to find movies/shows!! A 2-step process involving ChatGpt and Criticker.

It can be difficult to choose an appropriate film to watch when using Criticker's ratings as the only reference. Just because it has a high PSI doesn't mean you're interested in watching a movie of that genre right away.
As a result, I thought of utilizing ChatGpt (I need to get into the routine of always considering ChatGpt as an option when I have a difficulty).
So I asked it how I should use it to find a movie I would really enjoy watching:

It answered:
1. Tell me your favorite movies: Share a list of your favorite movies or describe what you like about them. This will help me understand your taste and suggest similar movies that you might enjoy.

2. Mention your preferred genres: Tell me which movie genres you enjoy the most, such as action, drama, comedy, sci-fi, romance, etc. This will allow me to suggest movies within those genres that have been well-received by audiences and critics.

3. Specify actors, directors, or other film personnel: If you have favorite actors, directors, or other film industry professionals, let me know. I can recommend movies they've been involved in, which might increase the chances you'll like the recommendations.

4. Define specific themes or elements: If you're looking for movies with particular themes or elements (e.g., strong female leads, plot twists, or coming-of-age stories), let me know. I can suggest movies that fit your criteria.

5. Request movies from a specific time period: If you prefer movies from a particular era, such as classic Hollywood or the 1980s, let me know. I can recommend highly regarded movies from that time.

6. Consider your mood: Tell me what kind of movie you're in the mood for (e.g., light-hearted, thought-provoking, or thrilling). I can tailor my recommendations to match your current mood.

Point 1 doesn't require a lot of time. Just go into your favorites and mark all the highest scores and copy/paste everything into ChatGpt. Afterwards, specify the genre and atmosphere you would like it to recommend. Also what you DON’T want e.g. horror, anime, documentary….
You can then evaluate the recommendations in Criticker. And be surprised to see that all of its suggestions have high PSI! Most of the 10 movies it suggests might be movies you already have seen. But then just ask for more and adjust the parameters along the way. I had to tell it that some suggestions were too cynical. That caused it to swing to the opposite direction, suggesting almost family movies. And in this way you can keep correcting if that is necessary.

I discovered the movie for tonight and 10 movies/shows on my watch list that I'm looking forward to watching in no time!

[Bryce74]

I tried ChatGPT too, for recommendations, but it didn't work for me. So I will continue with Criticker which gives much more personalised recommendations and I will continue to compare with IMDB ratings. I had hoped for more with chatgpt but it was quite disappointing. Even after a lot of talking to it, recommendations did not get much better. Even after feeding it my top rated criticker movies and series.

[NIV7]

Okay but you should stop using AI until the Earth is safe

https://www.theguardian.com/technology/ ... -emissions

I only linked one article but c'mon, read some news please.

[Scholar69420]

Sam Altman says that AI may ultimately require up to 10% of global energy demand to power the compute needed. However, if it helps invent new technologies that benefit the planet then it'll be worth it. The cat is out of the bag, nobody is going to stop using AI.

[VinegarBob]

I've been messing around with MetaAI the last few days. I came up with a novel way - pun intended - of getting recommendations from it. During a casual conversation about movies, books and music I told it my top ten favorite novels then asked it to recommend 10 films based on that list. It came up with an interesting batch of films, some of which I hadn't seen, and one of which I just watched today. I thoroughly enjoyed the film.

Chatting with an LLM as if it were a human friend is a fascinating and pretty fruitful way of generating all sorts of recommendations. 10/10 would recommend.

[Scholar69420]

Yup, the idea of AI performing in a social role is very interesting, especially with the personalisation LLMs offer. Instead of telling it my preferences I uploaded the CSV export of my criticker ratings, it even grounds discussion in the context of my mini-reviews.

[fringe]

I don't think that's gonna work.

[amazedemon]

You got me curious, so I shared my csv (feature films only) with an AI and asked it a few questions about average movie ratings per release year. These were the key insights.

Key Insights:
1. Recent years (2012-2022) have significantly more ratings than older years. This could be
due to more active users rating recent movies or a bias towards rating newer films.
2. The average scores for these recent, highly-rated years are lower than what we initially
saw for some older years. This suggests that when we have more data (more ratings), we
get a more balanced view of movie quality.
3. There's considerable variation in scores within each year, as shown by the box plot. This
indicates diverse opinions about movies released in the same year.
4. The year 2012 stands out with the highest average score among these recent years, while
2020 has the lowest. This could be interesting to explore further - perhaps 2012 was
considered a particularly good year for movies, while 2020 (the start of the COVID-19
pandemic) might have affected movie quality or viewer perceptions.

This analysis provides a more reliable picture of the ratings, as it's based on years with a
substantial number of ratings. The initial high scores we saw for years like 1947 were likely
outliers based on very few ratings, which can skew the results.
Would you like to explore any specific aspect of this data further?

I then went further down the rabbit hole trying to get it to make recommendations based on films I hadn't watched, but it faired less well. The recommendations were mostly films I had already seen, but not in my top 10.

I have a pdf with the conversation and python code it generated, but not the charts and tables, if anyone's interested.