ChatGPT: God? Precog? Skynet?
David Knauer
Managed Data & Security Services | Business Manager | Security Operations Center | Plan, Build, Run & Service | Data-Centric & Automation | #NewSchoolofSecurity
???????????? ??????'?? ???????????#CHATGPT????????? ??????, ????????????, ???????????? ?????? ?????????????? ???? ?????? ???????? ????
??????'?? ?????????? ???????? ?????? ?????????????????? ?????? ???????? ?????? ?????????????? ??:
The name itself probably disappoints already the most wet dreams of all the buzzword marketers in the world since they mostly only know the abbreviation of GPT:
?????? = ???????????????????? ??????-?????????????? ??????????????????????
Yes, pre-trained. No Skynet, God, Precog or Matrix behind it. Sorry for that. But it just does not sound that cool like fancy AI.,I know.
1. ??????????????: We don't know what ChatGPT is telling us ... it is pre-trained by a third party and it can give us biased information without even noticing
2. ??????????????: Why is it for free? Yes, because your data is the payment. No third party like OpenAI (even when the name is very philanthropic) is your friend and does something for free on the long term. Your data is used for sure. You just don't know for what.
3. ???????????????????????? ????????????????: Do you know where the data of your essay is coming from? Probably not. Do you know if there is something protected in your text? Do you quote the sources? Probably not. So in fact: You are violating IP. I hope you already feel like a gangster. You definitely will when the first lawyers notice they can easy make money with you.
4.?#????????????????: Let's skip those AI science fiction bull*hit, dangers of it and so on. But let's face the facts: There is no security regulation in ChatGPT.
5. ??????????????????: There are no daily updated information on ChatGPT like on Google.
6. ???????????????? ????????????????: Strongly related to 1. Google is useless when you don't know what to google. It multiplies with ChatGPT. It can be very dangerous when people ask the wrong questions without knowing and think they get the right answers because they don't do any research by their own? It is a real jeopardy to our critical thinking.
??????????????????:
Is ChatGPT cool? Yes, it is. Is it handy? Sure.
Should everyone use it like nature-given? No, they shouldn't.
ChatGPT has also many disadvantages and threats.
When people are the vulnerability the disadvantages and threats become risks.
So think about and manage your risks and don't follow the innovation marketing coach bull*hit train to imagination land.
?? ?????????? ???????? ?????????? ???????#CHATGPT??????????????? ?? ?????? ?????????????? ???
Last week I said my opinion about ChatGPT. Some people asked for more input and if I could advise usage without legal issues etc.
The good news: Now I can do.
The bad news: No, you can't use it right now with 100 % no legal issues.
The details:
1. ?????????????? ???? ???????? ?? ??????????????. It is not super smart. It cannot do anything bad without people. It just got trained by a huge amount of data and with a very impressive language model.
????????? ????????????????:
???- The database is not live (latest 31 of December 2022)
???- Nobody know what database is used
领英推荐
????? ????????????????:
???When you are fine with it, it is not an issue to use it. For many use cases you don't need a daily updated database. Plus, a live updated database will follow with an integration in Bing.
2. ?????? ?????????????? ?????? ?? ?????????? ??????. For me this one is the show stopper. I mostly work in a field where I need sources and people ask for proof. ChatGPT doesn't deliver it.
????????? ????????????????:
???- When you ask for proven facts without a source and you re-use it you definitely have a legal issue
???- ChatGPT sometimes give you wrong answers but you don't know why (e.g. a historical date is not exactly correct)
???- Biases by default
????? ????????????????:
???Don't use ChatGPT for a scientific work, proven facts without double checking, etc. For mails, non-scientific posts etc. it is totally fine. Challenge all the answers you get. Plus, I have heard the integration in Bing can show references and sources.
3. ?????????????????????? ?????? ??????????????. ChatGPT should work with GDPR in case of a right to be forgotten and it doesn't save the data in a personal structured manner (it is at least the statement by OpenAI).
????????? ????????????????:
???- For me unclear is if this is the truth ... because: You also can do reverse engineering with pictures to find original pictures which was used for the training ... for me it is not a right to be forgotten
???- In the European Union there will be the AI Act in near future and I don't have any idea how this is suppose to work with ChatGPT
????? ????????????????:
???OpenAI has to find a solution.
4. ???? ?????? ???? ???????? ?????? ???????????????? ????????????, ?????? ?????? ?????? ?????????????? ???????????????? ??????????.
???In my opinion ChatGPT won't bring cybercrime or?#cybersecurity?to a next level.
??????????????????:
ChatGPT is an impressive tool. It can be used for many things but scientific work etc. And be careful dealing with personal information. It is still a black box.
If you have any other questions: Just give me a comment.
?? I am David.
??? My passion is developing the new school of security.
?? I love to share my knowledge, especially to SME.
Cyber Solutions Developer and Architect at Trustwave - Microsoft MVP Azure Security
1 年The latest version of GPT-4 now incorporates image recognition capabilities. Consequently, if OpenAI logs are provided to the authorities, we could find ourselves in a 'precog' crime prediction scenario, almost as quickly as you can say "Bob's your uncle". Imagine a malevolent actor leveraging OpenAI to scan for potential victims, using Google Glass and crafting creative prompt engineering strings that tap into data from social media platforms such as Facebook. In this scenario, OpenAI can detect the attacker's behavior by monitoring its own logs and subsequently alert the relevant authorities. These authorities, equipped with Google Glass, can swiftly identify the offender within a crowd, gaining access to all the necessary evidence required to facilitate an arrest. Artificial Intelligence powerful to keep away from any government. This will likely hold true for a majority of organizations as well, especially when it comes to monitoring the usage of OpenAI on corporate hardware.