ChatGPT & AI: The Good, The Bad, and The Evil
A new Team member’s first article!
In today’s news cycle, it is difficult to miss all the fuss about AI, or more specifically, ChatGPT. So many differing opinions on the matter can make it hard to decipher what the future looks like. Few people think AI is a gimmick, but not many know the possibilities AI provides to us. According to Statista The immense number of users that flocked to ChatGPT is a show of its prowess and the curiosity that AI can instill into people. Whether an individual considers AI good or bad, the truth is it is here to stay, and will continue to evolve at an unprecedented rate. What does the future of AI look like? What about the current use-cases these programs have and how will they affect normal everyday life?
What is ChatGPT and why?
With its established relevance above, some may wonder what exactly ChatGPT is. OpenAI, an innovative AI research laboratory that was founded in 2015, and has received billions of dollars in funding, is valued at a whopping $29 billion. ChatGPT is a natural language processing chatbot driven by AI technology and models developed by OpenAI. ChatGPT is “fine-tuned” on top of their GPT-3.5 AI model that OpenAI trained. GPT-3.5 is a successor of GPT-3 that makes many improvements. It still has the same number of parameters (175 billion) but is not released as a standalone model and instead built into proprietary OpenAI applications and tuned for those specific use cases.
ChatGPT continues to prove itself in the eyes of the media. It has undergone numerous tests to evaluate its abilities. It is important to note that sentient AI is not close to becoming a reality, or even possible. With that being said, ChatGPT has IQ scores that range from low average to the 99th percentile. ChatGPT has garnered a myriad of achievements, some of which include passing an MBA exam, passing the bar in multiple states, and being incorporated into giant search engines for its ability to give information in a conversational context quickly. A quick overview of ChatGPT and AI makes it seem like the end-all be-all tool that will revolutionize the world, and while AI is certainly going to play a big part in our future, what are the actual current use cases and everyday applications this technology affords to regular people?
The good of ChatGPT lies in its ability to heighten productivity and expand one’s ability to attain knowledge. There is no doubt that many struggle to write well formatted, professional, and complex emails. ChatGPT is especially good at taking a person’s idea/statement and padding it out into a professional email. Your email could either be entirely written by AI, or you could give it an email you already wrote and ask it to improve the wording or check for spelling/grammar errors. If you do not like the way ChatGPT worded something or are going for a specific tone, you can ask it to change what it wrote for you. You do not need to copy and paste what it wrote back to ChatGPT, it will remember the context of the conversation, just like a human would. You ask ChatGPT questions and give it tasks just like you would to a person.
Another very helpful use case would be in analyzing/writing code. ChatGPT can either be asked to write a program from scratch, and generally will manage to succeed, or it can be given some code that a programmer wrote and then be asked to improve the code, explain the code, write comments for the code, or rewrite the code in a different programming language. In many ways ChatGPT is like if Google was a person. You do not take everything you see online as truth, and you shouldn’t with ChatGPT. It will sometimes get code wrong, or even blatantly lie about explanations of concepts and ideas. ChatGPT is in no way a replacement for a real human. AI compliments our abilities, allows us to write code faster, or be explained concepts and code in a way that is natural to us.
Another fascinating use case would be using it for legal battles and being explained laws. For many normal people it has always been difficult to decipher legal text. There are good reasons why law school takes so long. Though, what if there was a way for normal people to get legal consultation without the immense costs that come along with hiring a lawyer? As stated above, ChatGPT has passed BAR practice tests in many different states. DoNotPay is a legal services company that uses a chatbot to provide legal consultation to people. They advertise it as a “robot lawyer”. It can help contest parking tickets, cancel free trials, get refunds, sue people, attain visas, and green cards, and a couple other miscellaneous legal services. They claim to have a fairly high success rate with regards to contesting parking tickets, which is also a free service it provides. It does this all with the use of AI technology, the same kind the of AI that ChatGPT uses, but at a lesser scale and less advanced.
Possible use cases are seemingly endless. It can help students research, as it can recite sources, help create marketing material, write excel spreadsheet formulas, be used to analyze performance, perform complex calculations, and analyze network traffic or architecture and give feedback. It is certainly worth it to think over what you do on a consistent basis and ask yourself if it can be quickened or improved by AI. You might be surprised in the ways that AI can help your everyday life, but do not forget, using AI to its fullest is a skill in the same way Googling is.
Everything has its pros and cons, and the same can be said for AI/ChatGPT. The use cases AI affords may be indeterminable, but some considerations must be made before you use this technology. The main issue that arises when using AI is copyright. Who owns the data, media, essays, or ideas that AI can generate? In the case of ChatGPT, it learned from the internet. So, when you ask ChatGPT about a movie from the year 2000, it will know about it and be able to give you answers to your questions. But where did that answer come from? More than likely, a wiki, forum, or article on that movie. So, while the AI gave you the answer, there is an argument to be made that it is using others’ (potentially copyrighted) material. Another good example of copyright being an issue would be with code generated by AI. ChatGPT learned about programming from the internet, and as such others may have that code under an open license of some form. With ChatGPT being such a useful tool for programmers, it is difficult to say how it will be used in the future, or how copyright will be dealt with. This is an argument that needs to be settled in the courts, and as such, if you use AI for business purposes such as analyzing data or marketing images you will need to be sure that you are legally in the right. Currently that is impossible, as no one has gone to the courts with these issues yet, but it is certain to be more well defined in the future.
Another legal issue that could arise are data privacy laws. AI is, as established, is exceptionally good at analyzing patterns and data. Using it to look at customer data can be rewarding and lucrative. Though, if that data contains PII (personally identifiable information) or any other kind of protected data, it could be an issue. This only applies when using an AI service not made or managed by your own business. If you use an AI from another company, it can be difficult to know what they do with that data. Similarly, if you use an AI chatbot service, you must be careful with the information that chat bot collects and have a good understanding of how that chatbot processes the information, how the data/information is used, and where it is stored.
AI is objective and cold. It has no inherent ability to decipher intent. Morality must be instilled by the creator of the AI, not into the AI itself but by using a separate content filter for user input before it is used by the AI. ChatGPT can program mostly anything, and unfortunately that includes malicious programs as well. While OpenAI have implemented limits and attempted to stop people from using ChatGPT for malicious purposes, many have found ways to bypass these restrictions and get ChatGPT to generate malicious scripts and payloads. Not only can it generate generalized malware, but it can also be fed source code and then generate exploits for said source code, assuming vulnerabilities exist. This may seem to be worrying, but it is important to note that just like for regular programming use cases, this is a tool and not a replacement for people. Errors are possible and not uncommon. It still takes a skilled developer to take what AI generates and make it into something refined. So, when it comes to malware development, this will only speed up the process by which malware is developed. However, the most worrying use case may be AI generated phishing emails.
You know that AI can generate professional emails, but there is no way for it to confirm that you are who you say you are. AI simply generates the email that you asked for. Not only can you preform OSINT (open-source intelligence) using AI on a given company, with the context of conversation, you can then ask it to generate emails with the information it has gathered. This makes running phishing campaigns against a company much easier. Usually, we can detect phishing emails due to poor spelling and grammatical errors, but AI almost eliminates this issue entirely. Threat actors may now have a new and simple way to generate well formatted, grammatically correct, and spell-checked emails within seconds. This is where a good security posture for your company would be important. As phishing gets more sophisticated, so should your security. Thankfully, AI can be used to analyze network traffic and is already being used in EDR (Endpoint Detection and Response) solutions.
AI is an enormously powerful tool. It is up to the user to decide what use cases AI has for them, and how they use it, whether that be responsibly or not. There is a ton of information on the internet about hacking, but it is not restricted in any way. It will be the same with AI. Threats are going to evolve with AI, but so are the defenders. AI, while powerful is still a niche in most communities. Hopefully as the technology matures more will adopt its benefits before attackers do. Ultimately, AI has both heavy pros and cons. It ranges from very useful for everyday tasks to downright evil and malicious. Ideally by utilizing AI’s good, companies will be able to be more productive and stop threat actors quicker.
This article was written by a Genuine Human Data Security Analyst and Editors.
Original article by William Summers. Data Security Analyst, infotex