Camera reviews written by AI / AI Predicting Image Popularity

[jcs]

If yelp reviews written by AI can pass as real, so could camera reviews and social media posts (I've wondered about some posters on this site too ): https://news.developer.nvidia.com/ai-writes-believable-fake-yelp-reviews/. A great way to test AI would be on forums. So if someone is acting a bit odd, especially if they have a very anonymous presence- could be an AI / bot!

AI is also predicting image popularity for social media posts: https://news.developer.nvidia.com/ai-app-predicts-the-popularity-of-social-media-posts/.

Applying this tech to video would also be possible (much longer to train as well as to process once trained).

[iamoui]

Interesting. How do we know you're not AI/a bot?

[tomekk]

Hasn't this been known for ages now? Couple years back black hat seo had AI generating whole articles on all kinds of topics which would pass as legit for a lot of people. Fake reviews seem like a piece of cake for an AI in comparison. It seems like it's an easy article spinning AI software for reviews https://en.m.wikipedia.org/wiki/Article_spinning

 

 

[jcs]

3 hours ago, iamoui said:

Interesting. How do we know you're not AI/a bot?

LinkedIn: https://www.linkedin.com/in/jcschultz/. Perhaps contact some of the people who gave recommendations and ask if I am real.

The same question for you- what's your LinkedIn?

1 hour ago, tomekk said:

Hasn't this been known for ages now? Couple years back black hat seo had AI generating whole articles on all kinds of topics which would pass as legit for a lot of people. Fake reviews seem like a piece of cake for an AI in comparison. It seems like it's an easy article spinning AI software for reviews https://en.m.wikipedia.org/wiki/Article_spinning

I had not heard AI systems were so good they could pass 'mini-Turing' tests: so good the results passed as human.

It's understandable some people prefer to remain anonymous online. However for those anonymous accounts it may be challenging to convince folks that the account owner is human. For example, if an account simply disagrees with everything one says, that's a pretty simple 'mismatcher' program.

As the article points out, product reviews might become a challenge to find truth. Perhaps this is a new business opportunity to create trusted accounts connected to real people and authenticated with crypto. Or maybe we need to bring back brick & mortar stores for local shopping and encourage more local meetups/fellowship where people can discuss things face to face and/or in groups. I'd think Samy's in LA and similar places might host such events.

[iamoui]

1 hour ago, jcs said:

LinkedIn: https://www.linkedin.com/in/jcschultz/. Perhaps contact some of the people who gave recommendations and ask if I am real.

The same question for you- what's your LinkedIn?

That's exactly the answer I'd except from a bot/AI. 

As far as my LinkedIn, I don't have one. So, therefore, I must be a bot. 

[jcs]

43 minutes ago, iamoui said:

That's exactly the answer I'd except from a bot/AI. 

As far as my LinkedIn, I don't have one. So, therefore, I must be a bot. 

Live streaming video.

[PannySVHS]

 

Easy to manipulate or devide people on an online forum. Especially for AI, because it doesn´t know effort and never tires down. 

The communicational structure of an internet forum is very vulnerable to AI or other intelligent efforts with doubtful conduct.

So watch out for who you are paying attention to. The best thing to do is to appreciate my beautiful comments

AI will be hard to impossible to be recognized as such. My intelligence is so massive, so that some consider me the superiour AI and

neuro network connected to all other networks in the most meaningful way. I am god of bots, I am BOTTOGOD.

 

Data Analysis did not make the original joke work well, but Borats follow up. So it´s a Not Nooot Joke.

The original not joke is the content which modern dataanalysis has been providing us with for decades.

People swallow it and laugh about it. AI pushed this to perfection, premanufactured, remanufactured/reproduced content.

Always see the originator to see the original intention behind all the talking.

 

 

[iamoui]

1 hour ago, iamoui said:

That's exactly the answer I'd except from a bot/AI. 

As far as my LinkedIn, I don't have one. So, therefore, I must be a bot. 

Expect* not except. See, even AI make mistakes. 

 

Just now, iamoui said:

Expect* not except. See, even AI make mistakes. 

 

Wait, did I just reply to my own reply to point out my own spelling error? Am I becoming self-aware?

[jonpais]

Sorry jcs, I deleted all my social media accounts, but where can I find one of these bot gear reviews? Could you post a couple genuine ones and a bot one to see whether I can determine which is which? 

[BTM_Pix]

We have definitely had a few on here.

There have been three or four 'users' who popped up and offered some seemingly helpful yet clearly cribbed from a script information midway through a thread. They had been triggered by a camera model or term by the looks of it.

So, for example, someone might say

"Has anyone used the AF-C mode on a Fuji X-T2?"

and there would be a reply along the lines of

"In continuous-servo AF (AF-C), the camera continues to adjust focus to compensate for changes in the distance to the subject while the shutter-release button is pressed halfway. This makes it a good choice for subjects that are in motion." 

And there would usually be a link for 'further reading' about it. So the replies looked contextually correct but the information in them was always set out in this definition style. A few of them looked like they were triggered by a term from a user guide and then the section of the user guide was posted.

One interesting thing is that we seem to have not been invaded by the 'love and spirit guide guru' offerings that we used to get a lot of and I actually think they were triggered by the amount of discussion of the Black Magic cameras as the spambot often referenced black magic banishment in its offerings. Its definitely indicative of the decline in the discussion of the BM cameras for me.

Ironically, @jonpais I wonder if what you listed in the pinned thread regarding guidelines for posting on Al Jazeera might make this issue more prevalent over there as having a defined set of rules would make these bots more easily shaped to fit in with what a 'real' poster looks like? 

With regard to anonymity, it probably can make people more suspicious if you post under a pseudonym and without a real profile picture so I should probably remedy that myself and change my own profile name and picture to actually be me.

Best Regards

Mr. Pixtech 8500

 

[mkabi]

14 hours ago, jcs said:

LinkedIn: https://www.linkedin.com/in/jcschultz/. Perhaps contact some of the people who gave recommendations and ask if I am real.

 

I was telling jcs, how much alike we are in a PM after reviewing his profile.

Who else on this forum is like that???

[jcs]

11 hours ago, jonpais said:

Sorry jcs, I deleted all my social media accounts, but where can I find one of these bot gear reviews? Could you post a couple genuine ones and a bot one to see whether I can determine which is which? 

Hey Jon, as @BTM_Pix noted, the bot posts on this forum have so far been (mostly?) obvious. I just received that email from nVidia, so I haven't done any searching for more examples, specifically for gear/cameras/tech.

In college I wrote a program called ELMO (elementary linguistic motivator). It was a simple program that looked for keywords and common patterns that people could 'talk' to via keyboard. It would type responses with random delay, including making random mistakes (as @iamoui noted), backing up, correcting, and moving forward. It was entertaining to watch people interact with it (I took notes to make improvements for future revisions). So I can see the appeal developers have to test these systems on real people.

Through the massive power of GPUs and advances in machine learning, AI is now being used for world influence and for profit. Specifically for buying camera gear, here's the 'convincer program' (NLP term) I use before buying gear:

1. Learn about a new product, such as when Andrew does a blog post.
2. Read review replies on the forum.
3. Read reviews on Amazon and B&H.
4. Read reviews and watch videos from other bloggers/vloggers.
5. If all good, buy the product. I prefer to buy locally, such as from Samy's (the biggest camera purchase, the C300 II was from Samy's DV Edit). Low online prices are good, however at some point there will be little or no competition, less jobs for local folks, so hopefully more people will start buying locally when possible (even if it costs a little more).

Intuitively, when reading online reviews the detailed ones are more convincing, vs. generic "Great product!" which could be a bot or paid poster. As AI gets more sophisticated, detailed reviews could be bots. So the value of blogger/vlogger reviews with real people become even more important to find truth. @Andrew Reid and other gear reviewers could use this info to generate more traffic as people begin to realize what is happening to product reviews on sites like Amazon & B&H (and possibly even on forums as AI improves). As @tomekk noted, AI (and perhaps simpler statistics-based grammar engines) has been used for fake SEO.

Here are examples from the Yelp test, some are real, some are AI:

1. Easily my favorite Italian restaurant. I love the taster menu, everything is amazing on it. I suggest the carpaccio and the asparagus. Sadly it has become more widely known and becoming difficult to get a reservation for prime times.
2. My family and I are huge fans of this place. The staff is super nice and the food is great. The chicken is very good and the garlic sauce is perfect. Ice cream topped with fruit is delicious too. Highly recommended!
3. I come here every year during Christmas and I absolutely love the pasta! Well worth the price!
4. Excellent pizza, lasagna and some of the best scallops I've had. The dessert was also extensive and fantastic.
5. The food here is freaking amazing, the portions are giant. The cheese bagel was cooked to perfection and well prepared, fresh & delicious! The service was fast. Our favorite spot for sure! We will be back!
6. I have been a customer for about a year and a half and I have nothing but great things to say about this place. I always get the pizza, but the Italian beef was also good and I was impressed. The service was outstanding. The best service I have ever had. Highly recommended.

Answers here: http://www.businessinsider.com/researchers-teach-ai-neural-network-write-fake-reviews-fake-news-2017-8

Something to think about if a user is acting strangely, such as disagreeing with everything everyone says (there was a user like that here for a while- I thought maybe they were a bot, especially when I saw that user name on another social media site acting the same way).

Would be rather silly arguing with a robot designed for the purpose of arguing (unless one enjoys arguing, then by all means...)  

1 hour ago, mkabi said:

I was telling jcs, how much alike we are in a PM after reviewing his profile.

Who else on this forum is like that???

Right on brother, we're all alike more than we know. Namaste (for those unfamiliar- essentially means "I recognize me in you").

[jonpais]

17 hours ago, BTM_Pix said:

We have definitely had a few on here.

There have been three or four 'users' who popped up and offered some seemingly helpful yet clearly cribbed from a script information midway through a thread. They had been triggered by a camera model or term by the looks of it.

So, for example, someone might say

"Has anyone used the AF-C mode on a Fuji X-T2?"

and there would be a reply along the lines of

"In continuous-servo AF (AF-C), the camera continues to adjust focus to compensate for changes in the distance to the subject while the shutter-release button is pressed halfway. This makes it a good choice for subjects that are in motion." 

And there would usually be a link for 'further reading' about it. So the replies looked contextually correct but the information in them was always set out in this definition style. A few of them looked like they were triggered by a term from a user guide and then the section of the user guide was posted.

One interesting thing is that we seem to have not been invaded by the 'love and spirit guide guru' offerings that we used to get a lot of and I actually think they were triggered by the amount of discussion of the Black Magic cameras as the spambot often referenced black magic banishment in its offerings. Its definitely indicative of the decline in the discussion of the BM cameras for me.

Ironically, @jonpais I wonder if what you listed in the pinned thread regarding guidelines for posting on Al Jazeera might make this issue more prevalent over there as having a defined set of rules would make these bots more easily shaped to fit in with what a 'real' poster looks like? 

With regard to anonymity, it probably can make people more suspicious if you post under a pseudonym and without a real profile picture so I should probably remedy that myself and change my own profile name and picture to actually be me.

Best Regards

Mr. Pixtech 8500

 

We may come up with a set of community guidelines, but if we do, it'll probably be with input from Andrew Reid and the entire moderator team. Since you've uploaded your profile picture, now would be as good a time as any to remind everyone to please do the same if you haven't already. ? As far as pseudonyms go, they don't really bother me at all: of course, if you've got a great reputation, it doesn't hurt that contributors like Luke Neumann, Brian Caldwell, John Brawley and fuzzy normal all use their real names. ?