Written by Mirna, Ljubljana, 2023
It seems as if AI technology has already succeeded. Gaining attention from every tech workplace and professional online debate, it created an image in which ChatGPT has already changed the world for the better. But is that really so?
"The hype around #ChatGPT is justified, but we still have a long way to go in terms of developing truly intelligent machines." - Demis Hassabis, co-founder of DeepMind
Yes, it's fun, and sure, it has overcome some barriers, but OpenAI still faces many problems that haven't been solved. To move forward with being just a trend, ChatGPT has to address its limitations and risks first: 👇
They say it has everything, but at the moment it lacks a deeper understanding of the world, human feelings, or common sense reasoning. Large volumes of text data are used to train AI tools to recognize patterns and answer our questions. But they still may be unable to comprehend the finer points of human speech and language, though.
For instance, ChatGPT could find it difficult to comprehend irony, figurative language, or sarcasm (if changing prompts to more logical, it can show improvement).
And there’s a reason. Complex language models are only able to analyze info logically and systematically. They are not able to feel emotions, but they can recognize patterns in data that may signal specific emotions, such as facial expressions or speech tones. In other words, AI lacks awareness and feelings; it is not subject to the experiences of happiness, sadness, or rage.
We have to mention the obvious improvement in this area. While some prompts went viral about funny or confusing answers this AI tool provided, they made significant corrections in about 2 weeks. When asking a cultural or leaning towards inappropriate questions. ChatGPT no longer falls for it anymore. Try it yourself - check out the prompts used in January on this GitHub LLM failure archive and type them in today. And there you’ll have proof that the AI is learning about us, while we are learning about it.
As with any AI system, ChatGPT's responses can be biased if the training data contains biases. OpenAI has taken steps to mitigate this, but it still remains a limitation. Human error, sample bias, societal and historical influences, and other variables can all inject biases into the data.
For example, Meta shut down Galactica, an LLM (language learning model) that was comparable to ChatGPT, and trained on 46 million text samples, after 3 days because it spread racial and misleading information. It’s the reason everyone worries about AI being open to the public.
At a conference in Paris, Google just presented its new AI strategy. A new AI assistant called Bard, which is intended to be more human-like and combine with a range of goods, is the main focus of the plan. It runs on a versatile and adaptive AI engine, Google Assistant Pro. Google also intends to make its AI technologies available to other developers. Other AI-related announcements at the conference included new AI-powered healthcare solutions and a Paris-based AI research center. With a dedication to ethical and accountable AI technology, the significance of ethical issues in AI research was underlined.
Technology can be corrected, identified, and addressed by using methods like bias correction and data pre-processing. But until then, it is important not just to just take AI-generated answers for granted for now. When sources are not listed you should consider how to use your new AI answer.
Because ChatGPT isn’t providing citations, article authors, scientists, etc., don’t receive credit for their effort at the moment. And as ChatGPT could be a great content-generating tool, assisting companies in developing search engine optimization (SEO) plans. Businesses can utilize the AI tool to create high-quality, targeted content that is optimized for important target markets and top search engine keywords. But keep reading to find out why it is still far away from replacing websites.
The AI model lacks clarity regarding how it operates, tho not yet trustworthy or listing sources. How can we ensure that ChatGPT is being used for the greater good?
The risk of damage rises when AI systems get more complex and are applied to increasingly crucial tasks like medical diagnosis or self-driving automobiles. Incomplete design or usage of an AI system might result in mishaps or harmful judgments.
An example of real-world harm could present the system of an autonomous vehicle failing to recognize an impediment or interpreting sensor data incorrectly, self-driving cars might end up in an accident. Similarly to this, if an AI-powered medical diagnosis system is not trained on a variety of data, it may incorrectly identify a patient.
More lightly - the frequency of fake social media accounts is a major issue online, and the use of AI chatbots would make it simpler to commit online fraud. Another issue is the dissemination of false information, especially in light of ChatGPT's ability to make not-so-obvious incorrect responses appear to be correct. It’s a bit ironic that we have to verify we are humans before using it, and not vice versa. 🙂
This is why many people also worry about how AI may be used in cyberattacks or as a surveillance tool. All future AI development must work to guarantee that artificial intelligence technologies are created and deployed in a manner that prioritizes safety and reduces risks. Currently is a great time to start integrating and leveraging AI, maybe even building micro SaaS products, that can use ChatGPT through API.
Is it already taking over in 2023? We say, not quite yet but stay alert.
"The hype around #ChatGPT is both a blessing and a curse. It raises awareness about AI, but it also creates unrealistic expectations." - Yoshua Bengio, Turing Award-winning computer scientist
You can run into speculations like “Can ChatGPT take the place of Google as your preferred web search engine?” Although ChatGPT can theoretically be used for search, that is not how this particular version has been created. For starters, it cannot search the internet for the information you need and tell you where it came from because it does not crawl the internet. Second, the text was educated on its knowledge base, which ends in 2021.
The hype cycle can be best explained by the latest primary conclusions of the 2022 Gartner Hype Cycle for Emerging Technologies report, which assessed the maturity, business benefit, and potential of various technologies. (published in August 2022) The report provides insights for organizations to use in making decisions about using emerging technology, and the study analyzed at-the-time and forthcoming trends. The highlight is on: immersive workplaces, digital twin of the customer (DToC), NTFs, AI-augmented development, Web3, Superapps, and more. Gartner's report also emphasizes how crucial it is to create trustworthy AI systems and how important it is for AI to be morally and responsibly guided.
Where on Gartner’s graph would you place ChatGPT at the moment? Source: Gartner.com
Since August 2022 there were some drastic changes, particularly with the emergence of ChatGPT as a more convenient entry point for its technologies. Predictions state that ChatGPT is quickie reaching the peak of inflated expectations. But there is nothing to worry about if we remember that all disrupting technologies had their ups, downs, and stable stagnations before we merged them into our daily life. Today, developers are already building on ChatGPT SaaS products and not waiting for the second Gartners phase.
If you’re too young, skip this and if you’re not - remember the skepticism around the internet around 2001 and the bombarding of emails and .coms just a few years later. Seems like the same story. Seems like the time to start your story is now.
The new trend is also seen as a threat to take people’s jobs - but can other products replace the “hyped-up” ChatGPT? 💼 With the rise of Saas (Software as a Service) platforms, it's possible that ChatGPT won't completely replace human employees, but rather will be integrated as a helpful tool in various industries. Some are even waiting for positions such as ChatGPT or AI consultant to open soon. But because we can’t predict who’s going to win the grand Open AI race, here are some products to keep an eye on in the future, as they can turn out to be more useful:
The most recent innovation from OpenAI, ChatGPT, offers a lot to be excited about. However, there are some issues that are worth discussing beyond their immediate usefulness. The tech industry is hyped because it is the first open-source model of its kind & all developers (large, small, or hobby) got access to this technology at the same time. Are you still in “awe” or more “oh” zone about it?
And until then, have fun exploring ChatGPT and let us know in which areas it helps you the most. Although there's a lot of hype around ChatGPT, it's essential to bear in mind that they are still far from flawless and have their limits. The growth of Saas platforms provides a promising avenue for businesses to integrate chatbot technology while still retaining human employees. But, we may use these technologies in a way that helps businesses and customers by being aware of their pros &cons. It will be fascinating to see what new Saas apps will arise from these promising chatbot tools, as the area of natural language processing continues to develop.
Read more in our past articles where we focused on: