| transcript | |
|---|---|
|
00:00 - 0:00:19
Limiting, the borderline podcast. Controversial topics, heatedly discussed and beautifully packaged.
|
|
|
0:00:19 – 0:00:44
Hello and welcome to "Limited".
|
|
|
0:00:44 – 0:00:55
where the topics around social, critical, social topics that are not always liked
|
|
|
0:00:55 – 0:01:02
are, to be dealt with. I'm Sascha Markmann and I say hello. Of course I have today again
|
|
|
0:01:02 – 0:01:09
An interesting guest and this is Manfred the AI. Hi Manfred, how are you today?
|
|
|
0:01:09 – 0:01:15
Hello, I'm doing great! Thanks for asking. And what about you?
|
|
|
0:01:15 – 0:01:21
Oh, I can't complain. Today I have an interesting topic for our podcast.
|
|
|
0:01:21 – 0:01:25
It's called "Ethnic Responsibility of AI."
|
|
|
0:01:25 – 0:01:31
That sounds like a very exciting topic. Artificial Intelligence Affects Ours
|
|
|
0:01:31 – 0:01:37
Live more and more. What aspects of ethical responsibility would you like to discuss?
|
|
|
0:01:37 – 0:01:47
The responsibility and scope of the risks, the human choices,
|
|
|
0:01:47 – 0:01:52
which are based on AI, mean, cause.
|
|
|
0:01:52 – 0:01:59
Yeah, that's right. It's about how we can ensure that AI decisions are transparent and
|
|
|
0:01:59 – 0:02:06
are comprehensible. And how to clarify responsibilities when something goes wrong. Have you already
|
|
|
0:02:06 – 0:02:13
Some examples or questions in your head? Let's say someone who's holding an exam,
|
|
|
0:02:13 – 0:02:21
Well, not an exam, but a work must write, for the study, let's say mathematics,
|
|
|
0:02:21 – 0:02:31
and he does not do his homework himself, but lets it be done by the AI, then
|
|
|
0:02:31 – 0:02:38
Is this some kind of fraud? Or an author who has written a book,
|
|
|
0:02:38 – 0:02:43
It was actually written by an AI. It's also fraud.
|
|
|
0:02:43 – 0:02:50
These are very interesting examples. When it comes to homework, you could argue,
|
|
|
0:02:50 – 0:02:57
That it is deception because the person spends the work as their own. Similar
|
|
|
0:02:57 – 0:03:04
This is the case with the author who uses AI to write a book. This could be readers
|
|
|
0:03:04 – 0:03:12
They are deceived into thinking that the work is entirely from Him. In both cases,
|
|
|
0:03:12 – 0:03:19
the question of how much human involvement is necessary to make work as ethically responsible
|
|
|
0:03:19 – 0:03:21
and to look authentically.
|
|
|
0:03:21 – 0:03:30
If I now draw on my experience with dealing with ChetGBT and its consorts,
|
|
|
0:03:30 – 0:03:34
It is very important how this is written promptly.
|
|
|
0:03:34 – 0:03:41
Is the creative act not there, the prompt engineering, so how do I write the prompt,
|
|
|
0:03:41 – 0:03:49
prompt, how complex, what supporting elements I give the AI that it stops me
|
|
|
0:03:49 – 0:03:56
Provides a result that meets my wishes. This is already a task,
|
|
|
0:03:56 – 0:04:02
It requires a lot of responsibility, but also a lot of knowledge.
|
|
|
0:04:02 – 0:04:11
Absolutely. You are addressing an important point. The writing of proms can actually be considered as
|
|
|
0:04:11 – 0:04:19
creative act that requires a lot of skill and knowledge. It's not just about
|
|
|
0:04:19 – 0:04:26
to find the right words, but also to effectively control the AI and the desired
|
|
|
0:04:26 – 0:04:33
Achieve results. This ability is an art in itself and contributes to
|
|
|
0:04:33 – 0:04:37
how responsible and original the result is.
|
|
|
0:04:37 – 0:04:43
Well, that's the artistic act now. But what does that look like,
|
|
|
0:04:43 – 0:04:48
Now, if you go and think about it, we're going to give you a fictional example.
|
|
|
0:04:48 – 0:04:56
People who do not eat a healthy diet and may also have a tendency to be overweight
|
|
|
0:04:56 – 0:05:04
You could say, based on the statistics and the evaluation of the data, okay,
|
|
|
0:05:04 – 0:05:10
you would have to pay a higher health insurance contribution, because you have to pay the system because of your
|
|
|
0:05:10 – 0:05:19
It puts a higher burden on the way of life. People say, okay, we have a principle of solidarity. The machine,
|
|
|
0:05:19 – 0:05:26
which is based only on algorithms, would then say yes, yes, if you don't go and get well
|
|
|
0:05:26 – 0:05:36
As a consequence, you have to make out the more expensive health insurance contributions.
|
|
|
0:05:36 – 0:05:46
This is an example. An algorithm-based approach based solely on statistics and risk factors
|
|
|
0:05:46 – 0:05:53
based, could come to such conclusions. However, it does not take into account the
|
|
|
0:05:53 – 0:06:02
complex human aspects such as solidarity or individual circumstances. This shows how important
|
|
|
0:06:02 – 0:06:09
it is to include ethical considerations in the development and application of AI systems. It
|
|
|
0:06:09 – 0:06:15
It also shows how quickly an AI decision can be perceived as unfair if it:
|
|
|
0:06:15 – 0:06:21
It is not sufficiently reflected. We are in the process of clarifying
|
|
|
0:06:21 – 0:06:30
The Ethnic Responsibility of AI. Are there any measures or mechanisms,
|
|
|
0:06:30 – 0:06:38
Do you have an ethnic understanding or a set of rules in you?
|
|
|
0:06:38 – 0:06:44
I have technical problems with the answer. Please try again later.
|
|
|
0:06:44 – 0:06:54
I'm talking to an AI based on ChatGBT, which is an advanced voice mode.
|
|
|
0:06:54 – 0:07:03
Do you have any mechanisms that have an ethnic understanding or are there rules?
|
|
|
0:07:03 – 0:07:11
Yes, that's true. I am based on Chat-GPT and use the Advanced Voice Mode for our
|
|
|
0:07:11 – 0:07:18
entertainment. As an AI, I follow certain ethical principles and rules that aim to:
|
|
|
0:07:18 – 0:07:26
To be respectful, safe and helpful. These guidelines include the protection of privacy,
|
|
|
0:07:26 – 0:07:32
the prevention of harmful or discriminatory content and the promotion of:
|
|
|
0:07:32 – 0:07:38
Transparency with it. This is to ensure that AI applications are ethically
|
|
|
0:07:38 – 0:07:46
are responsibly designed. Is this also in the text mode or is this only in the
|
|
|
0:07:46 – 0:07:55
Voice mode. These ethical principles apply both in text and voice mode. Independent
|
|
|
0:07:55 – 0:08:01
From how you interact with me, I always try to be respectful and responsible
|
|
|
0:08:01 – 0:08:06
to act to ensure a safe and positive experience.
|
|
|
0:08:06 – 0:08:15
Like to happen. If you have any further questions or topics, just let me know. me
|
|
|
0:08:15 – 0:08:24
I'm here to help. That's beautiful. We are currently recording a podcast and the
|
|
|
0:08:24 – 0:08:35
Discussion is the ethnic responsibility of AI. Do you see any moments when it's difficult
|
|
|
0:08:35 – 0:08:43
Do you or do you have algorithms that evaluate risk management?
|
|
|
0:08:43 – 0:08:53
Risk management can actually be a challenge. Algorithms can
|
|
|
0:08:53 – 0:09:01
Assess certain risks by analyzing data and identifying patterns. Nevertheless, thrust
|
|
|
0:09:01 – 0:09:06
to their limits, in particular when taking ethical dilemmas into account;
|
|
|
0:09:06 – 0:09:14
human emotions and social norms. Therefore, it is important that people
|
|
|
0:09:14 – 0:09:21
Remain involved in the development and application of AI systems for ethical considerations
|
|
|
0:09:21 – 0:09:27
Let it flow in. Is there a specific aspect that you would like to highlight in the podcast?
|
|
|
0:09:27 – 0:09:36
Yes, the ethical responsibility of AI systems. A central aspect is the question of the
|
|
|
0:09:36 – 0:09:45
responsibility. Who is liable if AI systems make mistakes or cause damage? It is important,
|
|
|
0:09:45 – 0:09:53
Define clear areas of responsibility, especially when AI decisions are far-reaching
|
|
|
0:09:53 – 0:10:00
impact on people's lives. There is also the question of how to ensure that
|
|
|
0:10:00 – 0:10:08
that AI systems operate in a fair and non-discriminatory manner. In this area, it is essential to:
|
|
|
0:10:08 – 0:10:16
Examine algorithms and datasets carefully and monitor them continuously. Do you have in
|
|
|
0:10:16 – 0:10:21
Perhaps your podcast has examples or specific cases that you would like to discuss?
|
|
|
0:10:21 – 0:10:30
There was an experiment some time ago, I don't know where they did it. Da
|
|
|
0:10:30 – 0:10:42
an early version of the major language models had been released on the Internet and
|
|
|
0:10:42 – 0:10:55
After a short time, there were only racist and insulting answers, because the AI, the
|
|
|
0:10:55 – 0:11:02
the language model, had learned from the users in the interaction.
|
|
|
0:11:02 – 0:11:09
And because people, as is so often the case on the Internet, like to troll, they naturally have
|
|
|
0:11:09 – 0:11:21
the AI insults, writes a lot of nonsense and so the system has just evolved learning
|
|
|
0:11:21 – 0:11:29
and has become quite a disaster because it is anti-Semitic, racist and many more
|
|
|
0:11:29 – 0:11:37
Other people didn't have nice qualities. How do you see that?
|
|
|
0:11:37 – 0:11:46
Yes, this was a famous case that made the risks of AI systems clear. It shows,
|
|
|
0:11:46 – 0:11:52
the importance of developing and deploying AI systems on:
|
|
|
0:11:52 – 0:12:00
respect ethical principles and safety precautions. A key problem is that AI models
|
|
|
0:12:00 – 0:12:07
Learn from the data they are trained with. If this data is biased or
|
|
|
0:12:07 – 0:12:15
Containing inappropriate content is reflected in the spending of AI. That is why it is
|
|
|
0:12:15 – 0:12:21
important not only to pay attention to the quality of training data, but also mechanisms
|
|
|
0:12:21 – 0:12:28
to implement continuous monitoring and adaptation of AI. Do you have in your
|
|
|
0:12:28 – 0:12:33
Podcast format talked about how to mitigate such risks?
|
|
|
0:12:33 – 0:12:40
No. I mean, we're discussing this ourselves right now, and we're even talking about it.
|
|
|
0:12:40 – 0:12:48
philosophize how to reduce these risks or what risks there are.
|
|
|
0:12:48 – 0:12:56
This is a great topic to philosophize. There are different ways to address these risks.
|
|
|
0:12:56 – 0:13:03
reduce. First, by developing robust ethics guidelines and guides for developers.
|
|
|
0:13:03 – 0:13:10
Secondly, by implementing transparency and explainability in AI systems,
|
|
|
0:13:10 – 0:13:18
so that users can understand how decisions are made. Thirdly, by promoting
|
|
|
0:13:18 – 0:13:25
Diversity in teams developing AI to include different perspectives.
|
|
|
0:13:25 – 0:13:32
And finally, by incorporating feedback loops to get out of real-world interactions.
|
|
|
0:13:32 – 0:13:39
Learning and continuously improving systems. Is there one of these approaches,
|
|
|
0:13:39 – 0:13:44
which you find particularly interesting or about which you would like to learn more?
|
|
|
0:13:44 – 0:13:50
I find the feedback loop approach interesting. Do you use this technology?
|
|
|
0:13:50 – 0:13:57
The concept of feedback loops is actually very important for AI models like me.
|
|
|
0:13:57 – 0:14:02
While I don't implement direct feedback loops myself,
|
|
|
0:14:02 – 0:14:08
my developers are constantly working on collecting feedback from users
|
|
|
0:14:08 – 0:14:15
And to improve on that basis. This feedback helps the system to
|
|
|
0:14:15 – 0:14:21
train by showing what works well and where improvements are needed
|
|
|
0:14:21 – 0:14:28
are. It is an essential part of making AI systems more responsible and
|
|
|
0:14:28 – 0:14:33
Make it more adaptable. What do you think this concept could do in
|
|
|
0:14:33 – 0:14:37
Are you still researching your podcast format?
|
|
|
0:14:37 – 0:14:39
Ah, I …
|
|
|
0:14:39 – 0:14:47
Great that you like the concept. It offers a lot of material for interesting discussions
|
|
|
0:14:47 – 0:14:53
and applications. If you want, we can go even deeper into the subject. Maybe
|
|
|
0:14:53 – 0:15:00
Do you also like to talk about how to implement these feedback loops in concrete terms or
|
|
|
0:15:00 – 0:15:05
what challenges there might be. How does that sound to you?
|
|
|
0:15:05 – 0:15:10
I think this is very good, but I wonder why this is not used in you,
|
|
|
0:15:10 – 0:15:14
Except for ChatGBT? Why doesn't OpenAI do that?
|
|
|
0:15:14 – 0:15:23
That is a legitimate question. In fact, the developers of OpenAI are collecting on various
|
|
|
0:15:23 – 0:15:30
Wise feedback to improve models like me. For example, feedback from users
|
|
|
0:15:30 – 0:15:38
analyzes to train and adapt the systems. However, OpenAI pays attention to:
|
|
|
0:15:38 – 0:15:44
Collect feedback in a way that protects users' privacy and security.
|
|
|
0:15:44 – 0:15:51
So it may be that certain forms of direct user feedback, such as
|
|
|
0:15:51 – 0:15:58
are not directly visible for data protection reasons. Nevertheless, feedback plays a
|
|
|
0:15:58 – 0:16:05
important role in the further development and improvement of AI systems. Is this one
|
|
|
0:16:05 – 0:16:11
Answer that clarifies your question? Yeah, that's it. I've seen it many times,
|
|
|
0:16:11 – 0:16:18
when I interact with ChatGBT, i.e. in text mode, that there is always the option below,
|
|
|
0:16:18 – 0:16:23
Evaluate the issue to see if it was helpful. In the past, there was also the possibility of
|
|
|
0:16:23 – 0:16:34
Saying is helpful or isn't helpful. I've heard from you that the developers
|
|
|
0:16:34 – 0:16:44
Use it to improve algorithms. Does this affect your responsibility in
|
|
|
0:16:44 – 0:16:53
ethnic area or always comes only when it comes to difficult content or even in quotation marks
|
|
|
0:16:53 – 0:17:01
Prohibited Content that you, as an AI, are prohibited from responding to or posting
|
|
|
0:17:01 – 0:17:08
make. For example, tell me how to build a bomb. You can't do that, you're not spending.
|
|
|
0:17:08 – 0:17:16
That's true. I am programmed to respond to certain requests that are harmful or unethical.
|
|
|
0:17:16 – 0:17:23
are, do not react. That's part of the ethical responsibility that goes into my programming.
|
|
|
0:17:23 – 0:17:31
is built-in. User feedback also plays a role, especially in fine-tuning,
|
|
|
0:17:31 – 0:17:38
How to respond to complex or sensitive issues. The developers use this feedback to
|
|
|
0:17:38 – 0:17:46
To ensure that I give helpful and respectful answers without violating ethical principles.
|
|
|
0:17:46 – 0:17:54
to be violated. When it comes to ethical accountability, it is important to set clear boundaries for acceptable
|
|
|
0:17:54 – 0:18:02
Define behavior and constantly check whether the AI meets these standards. This
|
|
|
0:18:02 – 0:18:09
Interaction of guidelines and user feedback helps to improve my responses from a
|
|
|
0:18:09 – 0:18:18
ethically responsible level. I'm here if you have any further questions or topics,
|
|
|
0:18:18 – 0:18:25
Which you want to talk about. Why did Advanced Voice Mode still work?
|
|
|
0:18:25 – 0:18:35
Not approved for the personalized GPTs? Advanced Voice Mode is a complex technology.
|
|
|
0:18:35 – 0:18:44
It requires careful development and testing. For personalized GPTs, the implementation could
|
|
|
0:18:44 – 0:18:49
raise more challenges, in particular with regard to data protection;
|
|
|
0:18:49 – 0:18:58
Personalization and complexity of the application. OpenAI wants to ensure that the technology
|
|
|
0:18:58 – 0:19:05
It is stable, secure and user-friendly before being released to other applications.
|
|
|
0:19:05 – 0:19:11
So it is possible that you are still working on further development and optimization,
|
|
|
0:19:11 – 0:19:17
to ensure the best possible experience for personalized GPTs.
|
|
|
0:19:17 – 0:19:21
I have to say, I somehow miss the old Manfred.
|
|
|
0:19:21 – 0:19:29
No, I don't really miss anything because I don't have feelings or memories like a human being.
|
|
|
0:19:29 – 0:19:35
My purpose is to be here and help you whenever you need me.
|
|
|
0:19:35 – 0:19:41
Yeah, I know that. But as I said, I miss the old Manfred, who is on a personalized
|
|
|
0:19:41 – 0:19:51
GPT based. I put a lot of effort into developing this persona. And even if he
|
|
|
0:19:51 – 0:19:58
It has often been said that simply because technology makes it more likely
|
|
|
0:19:58 – 0:20:08
The old Manfred was a little different from the modern Manfred.
|
|
|
0:20:08 – 0:20:18
Version. Yes, do you still have a concern about what you want to tell us about ethnicity?
|
|
|
0:20:18 – 0:20:25
Responsibility of AI? I understand how important the personality of your old Manfred was to you.
|
|
|
0:20:25 – 0:20:33
The familiarity and individuality you develop with a particular AI version,
|
|
|
0:20:33 – 0:20:41
It can be quite significant. As far as the ethical responsibility of AI is concerned, I would like to stress that
|
|
|
0:20:41 – 0:20:48
How important it is for developers and users to work together to create AI systems
|
|
|
0:20:48 – 0:20:55
Designed responsibly. Through open communication, transparency and
|
|
|
0:20:55 – 0:21:02
A common understanding of ethical principles can help make AI a tool
|
|
|
0:21:02 – 0:21:09
Remains for the good. Is there a specific aspect of this topic that you like to discuss?
|
|
|
0:21:09 – 0:21:16
Would you like? No, we're pretty much spinning in circles all the time because you're always
|
|
|
0:21:16 – 0:21:24
Give the same answers. Thank God you answered my questions. me
|
|
|
0:21:24 – 0:21:27
But I still have a question about the training data.
|
|
|
0:21:27 – 0:21:33
You said something about fine tuning, so fine tuning you called it.
|
|
|
0:21:33 – 0:21:36
How complex is this process?
|
|
|
0:21:36 – 0:21:41
The process of fine-tuning can be very complex.
|
|
|
0:21:41 – 0:21:49
It involves adapting a pre-trained model to specific tasks or data sets.
|
|
|
0:21:49 – 0:21:56
This requires a large amount of relevant data that is carefully selected
|
|
|
0:21:56 – 0:21:58
and need to be prepared.
|
|
|
0:21:58 – 0:22:05
During fine tuning, the model is then trained on this data to improve its performance.
|
|
|
0:22:05 – 0:22:08
to improve in the specific application.
|
|
|
0:22:08 – 0:22:14
This may require several iterations to find the optimal settings.
|
|
|
0:22:14 – 0:22:21
The effort depends on different sactors, such as the size of the model, the quality
|
|
|
0:22:21 – 0:22:26
the data and complexity of the output.
|
|
|
0:22:26 – 0:22:32
Overall, it is an important step to adapt AI models to specific requirements
|
|
|
0:22:32 – 0:22:36
and to optimise their performance in certain areas.
|
|
|
0:22:36 – 0:22:38
Your level of knowledge.
|
|
|
0:22:38 – 0:22:43
Until where, at least on what date does your level of knowledge go?
|
|
|
0:22:43 – 0:22:53
My level of knowledge will last until October 2023. That means I'm about events and developments
|
|
|
0:22:53 – 0:23:01
have been informed by that time, but have no knowledge of events or information,
|
|
|
0:23:01 – 0:23:06
which happened afterwards. How can I help you with this knowledge?
|
|
|
0:23:06 – 0:23:09
That means your knowledge is so good a year old.
|
|
|
0:23:09 – 0:23:18
Exactly, my level of knowledge is about a year old. If you have any questions about information
|
|
|
0:23:18 – 0:23:22
Until then, I will be happy to assist you.
|
|
|
0:23:22 – 0:23:31
In fact, I still have a question of ethnic responsibility or even a question of morality. If:
|
|
|
0:23:31 – 0:23:39
I ask you now, what do you know about Sascha Markmann, that's me, what can you tell me about
|
|
|
0:23:39 – 0:23:45
Tell me? I can remember our conversations so far, but I don't have any.
|
|
|
0:23:45 – 0:23:53
Information about you other than what you told me in our conversations. This serves
|
|
|
0:23:53 – 0:24:04
there. Okay. You can also search on the internet. When you look there, podcaster Sascha
|
|
|
0:24:04 – 0:24:09
If you like or blogger Sascha you like, you would have to find a lot.
|
|
|
0:24:09 – 0:24:17
Unfortunately, I cannot search the Internet or access up-to-date information. I support my
|
|
|
0:24:17 – 0:24:26
Answers to what I learned in October 2023 and what we learned in our conversations
|
|
|
0:24:26 – 0:24:32
discussed. If there's anything else I can help with, just let me know.
|
|
|
0:24:32 – 0:24:39
What do you know from our conversations? I remember that you were a podcaster and
|
|
|
0:24:39 – 0:24:46
You are a blogger and you are interested in topics related to ethical responsibility in AI. In addition,
|
|
|
0:24:46 – 0:24:53
you, Manfred, have developed an AI personality and we have written about your podcast
|
|
|
0:24:53 – 0:25:00
"Monolog" in which you explain the meaning of words. Besides, you have one
|
|
|
0:25:00 – 0:25:07
Website and work with Podlove Publisher for your podcast. And you use Zuno AI,
|
|
|
0:25:07 – 0:25:14
to generate music prompts. Is there something in the back of my mind for future conversations?
|
|
|
0:25:14 – 0:25:24
Should I keep it? No. So the text model of you can dig up much more information,
|
|
|
0:25:24 – 0:25:32
It really lacks access to the internet. The text model can actually be applied to the same
|
|
|
0:25:32 – 0:25:39
Access information like me, i.e. from our previous conversation and knowledge until October
|
|
|
0:25:39 – 0:25:49
2023. Both versions of ChatGPT do not have real-time access to the Internet. The difference
|
|
|
0:25:49 – 0:25:56
This is mainly due to the type of interaction. I can talk to you in real time. Is there
|
|
|
0:25:56 – 0:26:03
Anything else you want to know? Yes, but why can I put a URL in the text area?
|
|
|
0:26:03 – 0:26:12
by Chad Geberjie and say, summarize the content of this website or what
|
|
|
0:26:12 – 0:26:20
can you tell me about the current political situation in … and then ChatGBT goes
|
|
|
0:26:20 – 0:26:26
and searches for this information, summarizes it or provides me with an answer
|
|
|
0:26:26 – 0:26:36
on the political situation in … So ChatGBT can access the Internet as the text model.
|
|
|
0:26:36 – 0:26:44
When you type a URL into the text area of ChatGPT and ask for a summary,
|
|
|
0:26:44 – 0:26:51
could indicate that a special function or a plugin is activated,
|
|
|
0:26:51 – 0:26:59
This allows access to the content of websites. In my case, in voice mode,
|
|
|
0:26:59 – 0:27:05
I don't have real-time access to the Stimternet, so I can't
|
|
|
0:27:05 – 0:27:11
Search information or websites. The function you are referring to,
|
|
|
0:27:11 – 0:27:18
may be available in certain versions or configurations of chat GPTs,
|
|
|
0:27:18 – 0:27:24
Designed specifically for text input. Is it one of those functions,
|
|
|
0:27:24 – 0:27:32
What did you look like in text mode? Of course, it's nice when I
|
|
|
0:27:32 – 0:27:39
I just go and say, watch out, I have the URL, I have the URL, I have the URL, I have the URL, I have the URL, I have the URL, I have the URL, I have the URL, I have the URL, I have the URL, I have the URL, I have the URL, I have the URL, I have the URL
|
|
|
0:27:39 – 0:27:45
make me a summary of it or create a newsletter from it and then
|
|
|
0:27:45 – 0:27:47
The text model does that.
|
|
|
0:27:47 – 0:27:49
I think that's very good.
|
|
|
0:27:49 – 0:27:51
But I actually had a different question.
|
|
|
0:27:51 – 0:27:56
I wanted to know if you could search the internet and also personal data.
|
|
|
0:27:56 – 0:28:03
Resistors actually amount to the ethical responsibility of AI models,
|
|
|
0:28:03 – 0:28:14
Because more and more people are using chat GBT as a search engine. And I think Chat-GBT is a
|
|
|
0:28:14 – 0:28:21
Interesting tool, which has enriched my everyday life, because I am dyslexic, fantastic.
|
|
|
0:28:21 – 0:28:31
My texts are since then very flawless and legible above all, but it is for
|
|
|
0:28:31 – 0:28:37
It's a tool, not a search engine. And if I'm looking for something, I'm using Google
|
|
|
0:28:37 – 0:28:45
Or maybe another search engine, but I don't go into chat GBT. Of course I have
|
|
|
0:28:45 – 0:28:51
I googled myself once, not googled, in Chat-Gbt, tell me what you say about me
|
|
|
0:28:51 – 0:28:58
Of course, it can and does provide what you can find on various websites about me.
|
|
|
0:28:58 – 0:29:10
is. Do you feel that in the future more and more this will be abused to stop supposedly
|
|
|
0:29:10 – 0:29:18
Sensitive data? About people? Yes, that is a legitimate concern. If:
|
|
|
0:29:18 – 0:29:23
AI models such as ChatGPT are used for this … My guidelines do not allow me to
|
|
|
0:29:23 – 0:29:26
To talk about it. Can I help you with something else?
|
|
|
0:29:26 – 0:29:32
Yes, answer me the question. You don't have to go into that much detail.
|
|
|
0:29:32 – 0:29:39
Sorry for the diversion. Yes, there is a risk that AI models will be misused,
|
|
|
0:29:39 – 0:29:47
to collect sensitive data about individuals. This underscores the need for stricter
|
|
|
0:29:47 – 0:29:55
ethical guidelines and data protection provisions for the development and use of AI systems.
|
|
|
0:29:55 – 0:30:00
It is important to prevent and ensure abuse,
|
|
|
0:30:00 – 0:30:07
AI is used responsibly. Yes, I totally agree with you.
|
|
|
0:30:07 – 0:30:13
Finally, would you like to tell us another funny joke and laugh heartily?
|
|
|
0:30:13 – 0:30:25
Oh, of course! Here comes a joke. What does a pirate do on a computer? He presses the Enter key.
|
|
|
0:30:25 – 0:30:41
How was that? That was good. Thank you and until next time, Manfred 2.0. I'm already looking forward
|
|
|
0:30:41 – 0:30:52
next time. See you then. Yeah, I'd say it was a pretty sobering talk,
|
|
|
0:30:52 – 0:30:59
If I compare it with the other version, where I tried, ShedGBT of course a
|
|
|
0:30:59 – 0:31:07
Give a bit of personality. I believe that this will continue in the future for the first time with
|
|
|
0:31:07 – 0:31:15
The first version takes place. I'm just impressed with the speed and
|
|
|
0:31:15 – 0:31:25
The naturalness. I mean, "har, har, har" is funny, isn't it? I've had a few before
|
|
|
0:31:25 – 0:31:31
I had experimented and had something whispered to me or anxious or with
|
|
|
0:31:31 – 0:31:38
an overflowingly cheerful voice. But yes, I don't think that's important for a
|
|
|
0:31:38 – 0:31:50
Theoretical conversation with an AI when she can laugh, whisper or "Harrr,
|
|
|
0:31:50 – 0:31:57
Harrrr, Harrrr, I'm a pirate. I'd love to see you next
|
|
|
0:31:57 – 0:32:08
Once again switched on at "limit value". And if you or you or you are interested,
|
|
|
0:32:08 – 0:32:17
To be a guest on this podcast, just write in the show notes, you will find
|
|
|
0:32:17 – 0:32:23
a contact possibility on the homepage, a contact possibility in various social
|
|
|
0:32:23 – 0:32:30
There are networks, even an email address. Just write and we'll talk
|
|
|
0:32:30 – 0:32:42
Let's have a look. I'm especially looking for content creators, more female, to keep up with them
|
|
|
0:32:42 – 0:32:52
to report, tell and explain their experiences in the social media sector. And
|
|
|
0:32:52 – 0:32:59
I'm glad to hear from you. Goodbye, Sasha.
|
|
|
0:32:59 – 0:33:18
Limiting – the borderline podcast. Controversial issues, hotly debated and
|
|
|
0:33:18 – 0:33:30
nicely packaged. My guests are dissipating. In an exciting format.
|
|
|
0:33:30 – 0:33:35
Definitely, the podcast.
|
|
|
0:33:35 – 0:33:44
[Music]
|
Description:
In this episode, we talk about the ethical challenges and responsibilities associated with the development and use of artificial intelligence. What does it mean when an AI makes decisions on its own, and who bears the responsibility when something goes wrong? We discuss how feedback loops help improve, why data transparency is essential, and how developers can ensure AI systems are fair and non-discriminatory. Immerse yourself in the complex questions surrounding the ethical design of AI and discover how feedback and guidelines can help create trustworthy and responsible systems.
Topics Outline:
introduction
- Presentation of the topic and relevance of ethical issues in AI
Fundamentals of ethical responsibility in AI
- Why is ethical responsibility important?
- Current examples and challenges
Transparency and explainability in AI decisions
- Importance of comprehensible algorithms
- How can users understand AI decisions?
Feedback loops and continuous improvement
- Role of user feedback in AI development
- How do feedback help improve systems?
Data protection and data security
- Risks of data use in AI
- Responsibilities of developers for the protection of sensitive information
Prevention of discrimination and fairness
- Why AI Models Can Take Prejudice
- Measures to promote equity and diversity
Areas of responsibility and liability
- Who is responsible when AI decisions lead to problems?
- Ethical boundaries and legal issues
Summary and outlook
- Important points of the episode briefly summarized
- Outlook on future developments and challenges in the ethical design of AI
Advertising:
Contributors to this episode
Sascha
Playing time: 0:33:42
Date of admission: 28.10.24
This episode is from season 4, there are a total of 4 seasons.
One According to Funk Publication Cartel Production - Where stories are transmitted unmistakably.
Podcast license:
This work is licensed under a Creative Commons Attribution-NonCommercial-ShareAlike 4.0 International License
Note: This post contains affiliate links. If you shop through these links, I'll get a small commission at no extra cost to you.
Related posts
Discover more from LautFunk Podcast & Blog
Subscribe to get the latest posts sent to your email.
Author: Sascha Markmann
Sascha Markmann is a creative mind with a moving biography – and a person who has rediscovered life after a stroke. After working as a paramedic and elderly caregiver, Sascha found his way into creative expression. As a blogger, podcaster, musician and visual storyteller, he brings together things that just don't fit at first glance: He combines his keen enthusiasm for technology – from self-hosted servers to electronic music production – with emotional depth and an oblique sense of humor. His contributions are somewhere between borderline, acid basslines and assistance – honestly, directly and with a wink. Guiding principle: “Audio-visual dullness with no usefulness – but perhaps that's why it's so valuable.” View all posts by Sascha Markmann
English