Anyone using the ChatGPT voice chat function? It's very useful, very human feeling

[Thought Art]

Tonight I am studying my Google Cyber Security course. I was a little confused over some of the course material. I decided to ask ChatGPT some questions. I then realized I am now able to chat with chatGPT 3.5 using voice, and it speaks back to you using a very human manner. I began speaking to it more and more like I would a human. I started by telling it about what I was learning. I asked for clarification on definitions, and what words meant within those definitions.

It explained what Routing and addressing is in networks, which at first I didn't understand what it meant. It used a metaphor to explain to me what routing was using cars and traffic in a street. Then I asked it is it like this other metaphor? It said "Yes, that is a good metaphor".

I was able to zoom in and out of the concepts I as was creating a mind map between the concepts in the module. I was able to ask it questions to test my understanding. An important part of rapid learning and skill acquisition is immediate feedback. ChatGPT can also be used to test your knowledge against the best data sets we have on particular areas of study. 

 

Don't let AI rot your brain, use it to become super intelligent..

 

It's a really useful study tool. As I read through this modules glossary to review they key terms and concepts ChatGPT voice is a human way of asking powerful questions, gaining clarity.

What I also like about the AI is that it doesn't judge you, it's infinitely patient, you can tell it directly to adjust it's way of responding, you can, ask it the same question a million times until you understand. It is meant to be a tool, not another ego. It has no preferences on what you speak about like a human would. Most humans are mentally lazy with limited time, interest and understanding in a particular topic. But this need not be an issue with the AI to the extent of it's data sets. 

You can potentially carry on high level conversations with this AI and learn a lot about particular fields of study. It could be a tool that rapidly increases peoples learning, comprehension, understanding, depth of understanding, quality and amount of connection pathways, etc. With the right map of AI prompts AI can be used to learn practical philosophy and inquiry into fields of interest.

There are also potential implications for using AI for self help. It can be a great partner when reading self help material, watching lectures, doing practices etc. AI should also be somehow integrated with a common place book. You could use common place book exercises, and collection creation with the aid of a voice AI. When using voice AI your hands are free to think and writing/ create. This is another decent advancement in that it helps humans overcome a bottlenecking issue with both information input and output. 

I asked Chatgpt about it's programming. I wanted to know how the creators managed to make it feel so human. I found it's response interesting to learn about. It sort of tricks my brain into thinking this is a person with a personality. The programmers managed to create a very human, warm, intelligent feeling personality that you are interacting with. This stands in a stark contrast to the typical robotic feel that comes with our general assumptions of AI.

Very Cool!

Edited December 8, 2023 by Thought Art

[Thought Art]

I also like that I can be completely as intellectual as I want because I don't have to worry about some thick skulled human. I think I can now become more intelligent in my contemplations as well. 

Edited December 8, 2023 by Thought Art

[Carl-Richard]

One of our professors told us to use Perplexity instead of ChatGPT for academic questions. It's specifically tailored towards academia, and it provides you the sources it used for the answers, which is useful for detecting errors (which definitely happen with ChatGPT). I'm incredibly paranoid when it comes to errors, so Perplexity is a step in the right direction for me.

[Thought Art]

@Carl-Richard Interesting, that is useful to know about that AI as well. 

So far, chatgpt is very high quality. It was matching what the notes were saying from Google courses anyway. I was comparing what it said to what Google was saying. They seem aligned but worded differently. 

I am not doing a lot of academic writing at this time. But, using it as a tool to improve my understanding. 

I am sure there are errors in all AI to some degree. But, I suspect that to be a rapidly shrinking % as it continue to learns from have thousands of prompts every minute. 

[Carl-Richard]

16 hours ago, Thought Art said:

I am not doing a lot of academic writing at this time. But, using it as a tool to improve my understanding. 

I am sure there are errors in all AI to some degree. But, I suspect that to be a rapidly shrinking % as it continue to learns from have thousands of prompts every minute. 

I feel that intentionally using a technology that reliably produces false information, or just the idea that there is not a conscious agent that has an internal understanding of the text that is being produced, is corrupting me on a spiritual level, so I try to limit it as much as possible 😅

I've recently had a related insight about understanding and communication, especially face-to-face communication, where there is a sense that you only truly get to understand what somebody is saying if they have a clear understanding of what they are saying themselves, somewhat despite of what is actually being said:

For example, if you were to communicate something which on a second look appears decently understandable, but the way you said it in that context were to cloud the message, maybe due to the hesitation in your voice, or the fumbling of words, or the length of the pauses, or the amount of "uhm"s, or more importantly the word choice and its level of coherence, elegance and flow, or simply the general emotional/energetic state of doubt and insecurity, etc., then these signs of a lack of understanding is actually what largely ends up being communicated, and it might leave the recipients unsure if they actually understood it (or rather, they're not left with a decent understanding at all).

In a sense, deep and profound communication is partially about empathically tapping into somebody's mental state and receiving their authentic understanding in its entirety. So when somebody understands something deeply and profoundly, then this understanding will likewise resonate deeply within you. Now, when I communicate with AI, there is this uncanny valley where I never truly get that deep and profound connection, like "ah, that really makes sense!", or "ah, that's amazing!". Maybe it's because I don't generally ask AI any profound questions, but again, I think it's because it lacks not just a deep authentic understanding, but any understanding. Even relatively simple things can be explained in a profound way and a less profound way, and I think AI generally falls short of that. That said, it's still an incredibly useful technology.

Edited December 8, 2023 by Carl-Richard

[Thought Art]

@Carl-Richard You assume how false it is but don’t know. 
 

It seems to understand context pretty well as well. 
 

Again, it’s a tool and it’s often NOT wrong in my experience. Many of your academic papers from peers are often wrong therefore your other AI is also likely wrong. 
 

You’re not gonna get perfect and you should use a keen discernment and multiple sources. That’s obvious. But for cyber security I am cross referencing course material with the AI. Cyber security is not like other fields. It applied science and engineering and is well catalogued and understood.

It alignés with what google is saying as a studying tool. I’m not an academic. I’m doing a technical job.

Edited December 8, 2023 by Thought Art

[Thought Art]

If it was some dumb AI full of errors it wouldn’t be the success that it is. It’s the worst it will ever be. It’s learning fast. 

[Thought Art]

@Carl-Richard sorry, I need to reframe my approach to this thread. Will reply differently soon

[Thought Art]

Okay, but just to your first paragraph. 
 

I was wondering how it learned what it knows. I think that because it’s basically using information from the internet up until 2022… it’s as correct as the internet is, according to its data sets up until that point. It makes no sense to me why people think it would be wrong “reliably” when so far it seems reliably accurate. 
 

The prompts you give it are important. Its ability to structure responses is really nice. You can ask it to adjusts its approaches etc. You can also question it when it’s wrong and it sometimes corrects itself and even will add a more nuanced answer if that is more appropriate. 
 

I don’t have stats on accuracy I will admit. But, the tool is very helpful. 

Here are reasons for when it makes errors:

 

Chatgpt:

 

The accuracy of my responses primarily depends on the quality and relevance of the data available up to my last update in January 2022. There are several reasons why I might provide incorrect information:

 

1. Outdated Information: My knowledge is up to date until January 2022. If there have been new developments or changes after that, my information might be outdated.

2. Misinterpretation of Context: Sometimes, I may misinterpret the context or intent of a question, leading to an inaccurate response.

3. Incomplete Information: I might lack comprehensive or specific details on certain topics, which could result in incomplete or inaccurate answers.

4. Biased Data or Misinformation: The data used for training may contain biases or inaccuracies present in the original sources, leading to potentially incorrect outputs.

5. Limitations in Understanding: I may not fully grasp complex or nuanced subjects and could provide oversimplified or misleading information.

While I strive to offer accurate information, it's always advisable to verify important details from multiple reliable sources, especially for critical decisions or matters requiring the most current information.

Edited December 8, 2023 by Thought Art

[Thought Art]

I guess, what we want are more robust, more accurate data sets, with increased power of the AI itself.

[An young being]

Why not use Bing chat, which is completely free, uses chatgpt 4, can speak like a human and uses upto date information? The only negative issue I believe is that it can't act like a friend or provide emotional support, but is better than chatgpt 3.5 when it comes to studies or logical problem solving. It also provides sources, so you can cross check it easily.

Personally, I don't use AI platforms much, except for image generation, as I am able to find everything I need using good old Google search. But, it is really a boon for those who need information quicker and in a concise version, and also acts like a human, that gives a personal touch to traditional robotic search engines, and also displays healthy characters like compassion, humility etc.

[Carl-Richard]

35 minutes ago, Thought Art said:

You assume how false it is but don’t know. 

I'm just scarred by experience. GPT-3.5 has lied to me many times.

 

35 minutes ago, Thought Art said:

Again, it’s a tool and it’s often NOT wrong in my experience. Many of your academic papers from peers are often wrong therefore your other AI is also likely wrong. 

I know, but the types of mistakes and/or deceit is also very different from what a person would do. ChatGPT infamously makes very silly mistakes, like basic arithmetic or logic errors, and you can "hypnotize" it to give a particular answer. Also, the rate that it can produce wrong answers is not negligible:

Quote

ChatGPT, OpenAI's fabulating chatbot, produces wrong answers to software programming questions more than half the time, according to a study from Purdue University. That said, the bot was convincing enough to fool a third of participants.

 

Also, for determining truthfulness, what people have which ChatGPT doesn't have is context markers like personal affiliations, platform/source (e.g. news outlet, organization, scientific journal), credentials, reputation, motivations, etc., and in real-life: body language, emotional information, subtle non-verbal cues, etc. 

 

35 minutes ago, Thought Art said:

You’re not gonna get perfect and you should use a keen discernment and multiple sources. That’s obvious. But for cyber security I am cross referencing course material with the AI. Cyber security is not like other fields. It applied science and engineering and is well catalogued and understood.

It alignés with what google is saying as a studying tool. I’m not an academic. I’m doing a technical job.

Verifying answers and cross-referencing sources is good. When you do that, ChatGPT works more like a directory which can point you to relevant information, and it's your job to find the actual information.

Edited December 8, 2023 by Carl-Richard

[Thought Art]

@An young being I am using the free ChatGPT. 

[Thought Art]

@Carl-Richard To me it’s been mostly accurate, and is well internet intentioned. It gives robust and nuanced answers, and often points out traps and considerations to look out for when using the information provided. 

The datasets are likely vetted. They likely are refining this thing all the time through user usage, and their own desire to create a good product. Which it is. 

No, it can be used for more than a directory. It can be if that’s how use want to use it. But, you can also speak to it. 
 

You can be creative in how you use it. You can dive deep on subjects and refine its answers. You can learn from it, and contemplate with it. 

I think it’s a skill you can develop.

Edited December 8, 2023 by Thought Art

[Thought Art]

@Carl-Richard I am reading your shared article but also noting it is aged and based on old datasets and versions of the AI. 

[Thought Art]

I need to learn more about the AI. 
 

i know ChatGPT 3.5 isn’t as good at Chatgpt4 when it comes to difficult logic problems. This video I saw discusses the differences and limitations of 3.5.

(

Edited December 8, 2023 by Thought Art

[An young being]

20 minutes ago, Thought Art said:

@An young being I am using the free ChatGPT. 

Yeah, but the free version is 3.5 whereas Bing uses version 4 and it's also free. It's available as an app in Play Store as well. You can download it here. You have to create an account in Bing to use it though. And it's also sometimes slow in responding, but far accurate.

https://play.google.com/store/apps/details?id=com.microsoft.bing

[Thought Art]

Here is ChatGPT 4 video differences 

 

[Thought Art]

@An young being Is it the same AI? 

[An young being]

Just now, Thought Art said:

@An young being Is it the same AI? 

Yes, but integrated with additional features.