Chatting with ChatGPT...

threewins

Come across anything useful or interesting?

AnnPT77

Other than the guided meditation in Dolly Parton's style that someone asked it to create, no, not so far.

Why would a piece of AI, working the way that one works, create something truly useful, except by accident, a sophisticated variation on monkey-typewriter? (Unless someone's trying to cheat on their term paper: I guess that's a use.)

Interesting, sure. Maybe provocative. But it doesn't have any grasp of meanings, just patterns. Really nuanced ones.

refactored

It does mash up content it got from the web. It can be useful. I use it in my work and it has helped me solve various IT related problems.

Here I asked it how to lose weight ...

Losing weight involves making sustainable changes to your lifestyle, such as adopting a healthy diet and increasing your physical activity. Here are some tips that can help:

Start with small changes: Gradual changes to your diet and exercise routine are more sustainable than trying to make drastic changes all at once. For example, start by reducing your portion sizes or adding a few extra minutes of physical activity each day.

Eat a balanced diet: Focus on eating whole foods that are high in nutrients and low in calories. Fill your plate with plenty of vegetables, fruits, lean protein sources, and healthy fats.

Reduce your calorie intake: To lose weight, you need to create a calorie deficit by consuming fewer calories than you burn. This can be achieved by reducing your portion sizes, choosing lower calorie options, and avoiding high calorie foods and drinks.

Increase your physical activity: Exercise is important for weight loss and overall health. Aim for at least 30 minutes of moderate-intensity exercise most days of the week. You can also try incorporating more physical activity into your daily routine, such as taking the stairs instead of the elevator.

Stay hydrated: Drinking enough water can help you feel full and may reduce your calorie intake. Aim for at least 8-10 cups of water per day.

Get enough sleep: Lack of sleep can interfere with weight loss efforts. Aim for 7-9 hours of sleep per night.

Remember that sustainable weight loss is a slow process, and it's important to be patient and persistent. It's also important to talk to your doctor before starting any weight loss program to ensure that it's safe for you.

And then I asked should I cut my calories to 1200 per day to lose weight ...

Cutting your calorie intake to 1200 per day may result in weight loss for some people, but it's not necessarily appropriate or safe for everyone. The amount of calories you need to consume to lose weight depends on several factors such as your current weight, height, age, sex, activity level, and overall health status.

In general, a safe and sustainable weight loss rate is around 1-2 pounds per week. To achieve this, you need to create a calorie deficit of about 500-1000 calories per day, which can be achieved by reducing your calorie intake and increasing your physical activity.

Cutting your calorie intake to 1200 per day may be appropriate for some individuals who are smaller in size or who have a low level of physical activity. However, if you are very active, taller, or have a higher body weight, 1200 calories per day may not be enough to meet your nutritional needs and could lead to nutrient deficiencies, muscle loss, and a slower metabolism.

It's important to speak with a healthcare professional or a registered dietitian before making any drastic changes to your diet or exercise routine, to ensure that it's safe and appropriate for you. They can help you determine your individual calorie needs and guide you in creating a safe and effective weight loss plan that is tailored to your needs and lifestyle.

refactored

Here it is commenting on the keto diet ...

The ketogenic diet, or keto diet, is generally considered safe for healthy individuals when followed correctly. It's a high-fat, low-carbohydrate diet that aims to put the body into a state of ketosis, where it burns fat for fuel instead of carbohydrates.

However, the keto diet is not appropriate for everyone, and there are some potential risks and side effects that need to be considered.

One potential risk of the keto diet is nutrient deficiencies. Since the diet restricts many food groups such as fruits, whole grains, and legumes, it may be challenging to meet the body's nutritional needs. Additionally, the high intake of fat and protein may put a strain on the liver and kidneys.

Another potential risk of the keto diet is the "keto flu," a collection of symptoms that some people experience when starting the diet, such as headaches, fatigue, and nausea. These symptoms usually resolve within a few days to a week.

There is also some evidence to suggest that the keto diet may have negative effects on cholesterol levels, although more research is needed in this area.

It's important to note that the keto diet is not recommended for individuals with certain medical conditions, such as liver or pancreatic disease, or for pregnant or breastfeeding women.

Before starting the keto diet, it's important to speak with a healthcare professional or a registered dietitian to determine if it's safe and appropriate for you. They can help you create a personalized plan that meets your nutritional needs and fits your lifestyle.

cmriverside

^^Those are pretty dang good answers!

/forum

Sinisterbarbie1

I think it does pretty well where answers can be crowdsourced/culled from lots of data amassed from the internet such as the questions about diet above. From what I have read, and my anecdotal experience as a university professor, the challenges appear when questioners ask the AI to solve problems. The AI wants to be “right” and being “right” means repeating past patterns of success in its analysis of training data. In the examples I have read about the questioner on the other hand might be seeking to break away from old ways of hiring/calculating credit scores/determining bail flight risk or parole eligibility etc. = issues that may be infected with either direct or hidden bias or years worth of data of doing things in certain ways the questioner seeks to avoid perpetuating. For instance at one company the prompt was to select more diverse candidates to interview that will be likely to be successful engineers at the company. Nevertheless the AI returned mostly male and mainly white candidates- modeling its succesfull search after the characteristics in the profiles of the highest ranking engineers currently employed by the company . The company giving that prompt was unable to eliminate bias in the AI no matter what it told the AI to exclude, (don’t include gender, don’t look at race) because the AI had determined that white men were the most successful in the company and so kept looking for other proxies for those characteristics when told to ignore gender, race etc. (it looked to things like sports teams played on, fraternity/sorority membership, and even more subtle suggestions of gender and race apparently).

Another challenge I have noticed when experimenting to see how it would do on university exam questions we develop is that the AI tends to be repetitive in its response, and not have any/or accurate citations for facts it relies on, and to say things that sound good/plausible but are actually made up. So the answers are currently unreliable in many areas if a rigorous exam is constructed. I do think they are a useful tool for students to learn how to use - sort of like a calculator or a search engine can be a tool to aid learning and research, but needs to be used with a filter and one’s own logic and knowledge applied to it.

Finally, many authors and artists are raising ethical objections to the use of their works to train AIs by wholesale copying of books, art etc. into these neural networks that is then ‘mashed up’ and output as new works - sometimes even with the signatures of the copied artists attached — all without their authorization or approval, and without any compensation going to the artist (while the companies creating the AIs are amassing valuable data about users of the AI chatbots and using it for advertising and other privacy invading purposes the users probably aren’t yet considering).

Some of the avatar artwork i have seen is interesting, but others unduly sexualize women in particular where that has not been asked for, or they come out creepy in odd ways. One of my male colleagues had the subtly creepy experience … he received one avatar that made him look just like tom cruise, another where he was depicted wearing a short sleeved business suit (weird?) and a third where he had a big friendly smile but three rows of teeth. All were generated as a package from the same image and prompt to develop an artistic avatar from the photo to use on Linked In.

threewins

I've found ChatGPT useful for some things, and light entertainment for others. It told me why my eczema was not on the palm of my hand and gave a name to bursts of noise in my head when falling asleep ("Exploding head syndrome"!!!). It's kind of useful when someone posts a long unparagraphed screed, I just ask ChatGPT to give me a summary. I'm learning computer programming and one exercise it did produced code almost exactly what the tutor gave as a solution (a few years ago).

Some of the answers about culture can be totally wrong which makes me distrust it for other things as well. Quotes from films (even when it has access to the script) can be totally made up and when I'm trying to remember the name of a film its suggestions can be laughable.