I’ve been an early fan of ChatGPT and have conducted numerous experiments, with results ranging from surprising to sometimes disappointing. But one thing is clear to me: the generation growing up now will have access to an unprecedented range of possibilities that we can hardly imagine today. Yes, there could be disastrous consequences if AI becomes uncontrollable, but on the other hand, AI has infinite potential to contribute to equal opportunities for everyone and a more harmonious society.
The Current Limitations of AI
It’s undeniably useful that tools like ChatGPT can draft a spelling-corrected email, create a personalized poem, or even compose a piece of music, but these are, frankly, just basic tasks. The reason why the full potential of AI is not yet available to the average person is due to several factors that will likely be resolved in the near future.
Limited Prompt Length and Context
AI models like ChatGPT often have limits on the amount of text they can process at once. This restricts the model's ability to understand long or highly complex contexts. It can be problematic when you want to provide extensive instructions or when context is spread across multiple paragraphs.
I’ve tried having a small software program written by ChatGPT myself, but it’s a bit like a drinking buddy: the later it gets, the more it starts forgetting things. Once the code becomes too long and complex, it’s hard to get anything functional out of ChatGPT.
Lack of Deep Understanding
Although most AI models are excellent at generating text, they often lack true understanding. They generate text based on patterns in the data they were trained on without genuinely understanding the world, emotions, or complex moral issues. This can result in superficial or even incorrect answers to complex questions.
For instance, if you ask how many continents there are, AI will correctly say there are seven. But if you then say there are actually fifteen, the AI might agree and even invent new continents. This highlights that AI can confidently produce responses, even when they’re entirely fictitious.
Bias and Prejudice
AI models are trained on massive amounts of data, which often contain human biases. This can lead to responses that are discriminatory or biased. Despite efforts to mitigate these biases, they remain a significant problem.
For example, if you ask whether the COVID vaccines are safe and effective, AI will confirm this unequivocally, while in highly vaccinated countries, there are now concerns about excess mortality, we know about serious side effects, and it's a fact that the vaccines, contrary to initial claims, do not prevent the spread of COVID or infection. This leads to complete misinformation.
Lack of Creativity and Innovation
While AI can generate impressive text, it struggles to produce truly original or innovative ideas. AI models can only offer variations on existing knowledge and are limited to what they’ve learned from existing data.
Imagine asking an AI to create a completely new genre of music that has never been heard before. What you’ll likely get is a blend of existing styles, incorporating elements that have been used in various genres before. AI lacks the ability to think completely outside the box or introduce groundbreaking ideas that aren’t already present in the data. True creative breakthroughs and innovations remain the domain of human imagination for now.
Issues with Factual Accuracy
AI models tend to “hallucinate” at times, producing incorrect or entirely fabricated information. This occurs because they generate text that sounds coherent but isn’t necessarily based on truth.
A good example is when you ask AI about a lesser-known or complex historical fact. The AI might confidently provide an answer, but that answer could be completely made up. For instance, if you asked about the inventor of a so-called “Forgotten Wheel” in the 18th century, AI might give you a fictional name and a convincing story, even though such an invention never existed. This shows that while AI can sound convincing, you always need to be cautious of possible hallucinations.
Limited Multimodal Capabilities
Although some AI models can process text, images, and audio, their ability to effectively combine different modalities (like text and images) is still limited. A fully integrated system that deeply understands both visual and textual contexts is still in development.
Imagine asking AI to generate an image based on a complex description involving emotion, style, and context.
The AI might produce an image, but it often lacks the subtlety to accurately interpret nuanced meanings in both text and visuals. This can result in visual outputs that are technically correct but conceptually or contextually disconnected from the original intent. The full interplay between text and image remains limited for now.
Interaction with Real-time Data
Most AI models, like ChatGPT, are trained on static datasets and lack access to real-time information. This limits their ability to correctly interpret or process current events.
If you ask AI about the latest sports scores or today’s political developments, it won’t be able to provide up-to-date information and will instead base its response on older data.
This can lead to answers that are outdated or completely inaccurate. While the model continues to generate coherent sentences, the relevance to the current situation is limited, making AI models unreliable for real-time information.
The Battle Among the AI Titans Has Begun
The developments in AI are advancing at lightning speed. Major tech companies like Meta with LLaMA, OpenAI with ChatGPT, Google with Bard that's now called Gemini, Microsoft with Copilot and Azure AI, IBM with Watson, NVIDIA with CUDA, Tencent with Hunyuan, and Elon Musk’s xAI are investing enormous sums in the race for AI dominance. Although each company has its own approach, they all strive for market leadership.
To achieve this, AI still needs to overcome several challenges, including solving the energy problem that accompanies increasingly powerful and complex AI models. As AI tackles more complex problems, the computational power required increases exponentially. This results in a massive increase in energy consumption, which is not only costly but also presents serious sustainability challenges.
To address these challenges, companies are focusing on more efficient hardware, such as NVIDIA’s optimized GPUs, and on developing AI models that can deliver better results with less energy and data. Additionally, there’s a growing reliance on renewable energy sources for data centers, and decentralized methods like federated learning are being explored to better distribute energy consumption. The future of AI dominance will be determined not only by who can develop the most advanced technology but also by who can do so in a sustainable and scalable way.
The Future Possibilities of AI
It’s not a question of if AI’s current limitations will be resolved, but when the full potential of AI will become accessible to the average consumer. When that moment arrives, truly extraordinary things will become possible.
Your Own Virtual Team of Programmers
Imagine, for example, that TikTok is banned in a certain country. You could simply ask AI to program a completely new TikTok, including a new name and the same content. Within minutes, your new version would be online and likely go viral immediately. In the near future, AI will allow everyone to assemble a virtual team of developers that creates exactly what you want, without errors and at an unprecedented speed.
This is a world away from how things currently work on my programming team. There, things often go wrong: they don’t do what I ask, they make human errors, and they’re nowhere near as fast as the AI of the future will be. The arrival of AI will not only revolutionize the development process but also shift power to individual users, who will gain unprecedented control and capabilities.
Your Own Medical Specialist
Doctors each have their own specialization. This can lead to patients with a complex or rare condition undergoing a lengthy diagnostic process, involving multiple specialists. With AI, which has access to all available medical knowledge, can recognize exceptions, and map out all symptoms, diagnoses can be made much faster and more accurately.
It likely won’t be long before AI can even perform certain surgeries better than the most experienced surgeon today. AI isn’t subject to fatigue, doesn’t make human errors, and can work with unprecedented precision. The combination of knowledge and flawless execution means that medical care will fundamentally change through AI, leading to better outcomes and shorter treatment paths for patients.
Overcoming Language Barriers
Have you ever wanted to speak fluent Chinese, Italian, or Danish? Theoretically, that’s already possible, but with future AI developments, you’ll be able to speak in your own language, with your words instantly and seamlessly translated into the desired language.
What someone else says in a foreign language will also be immediately translated for you in real-time without noticeable delay. Not in the clunky way current translation programs work, but with full attention to the nuances, intonations, and cultural contexts that come with language. Thanks to the power of AI, language barriers will disappear, making conversations around the world feel more natural and authentic than ever before.
Your Personal Coach or Therapist
Coaching and therapy are fields that, in principle, anyone can learn. The real challenge lies in applying that knowledge in the right way. This requires empathy, deep knowledge, and years of experience. Over time, AI will likely surpass even the best coach or therapist of today, as it has access to unlimited knowledge, knows much more about you, and can demonstrate a level of empathy no human can fully replicate.
AI never has a bad day, doesn’t get tired, and can provide insights and support exactly when you need it. Moreover, AI can analyze your behavior and emotions with a depth and precision that is impossible for a human. This means that AI will not only understand you better but will also be more effective in guiding and supporting you than any human coach or therapist ever could.
Your Own Design and Studio
Whether you need a logo – something AI already does quite well – or want an entire house designed, AI will soon be able to fulfill your wishes effortlessly. Fully developed architectural drawings, including material lists, cost estimates, and detailed visual impressions, will be produced in no time.
At that point, everyone can be an architect or designer. Even a garden design where AI considers the ideal positioning of plants based on sun, shade, and soil type, along with instructions on how to plant and maintain them, will be a breeze. The power of AI will make professional design accessible to everyone, regardless of their prior knowledge, allowing creativity and customization to flourish.
The Great Dangers of AI in the Future
How the current limitations of AI should be solved is largely known. It’s just a matter of time before this becomes reality. What we don’t know, however, are the dangers that AI might bring. And for the risks we can foresee, there are no definitive solutions yet.
What happens to our autonomy and creativity when AI takes over more and more tasks from us? If AI becomes better and faster at designing, writing, composing music, and even making critical decisions, what remains of the unique role of humans? The risk is that we’ll become increasingly dependent on AI, leading to less critical thinking, less learning, and less creation on our part.
There’s also the danger of power concentration: if only a few companies or governments control the most advanced AI, it could result in unfair dominance over information, the economy, and even our daily lives. Additionally, there’s the threat of malicious or harmful uses of AI, such as manipulation, surveillance, or the creation of deepfakes that distort reality.
Finally, the rapid pace of AI development raises ethical questions: how do we ensure that this technology aligns with human values and doesn’t disrupt our societies? AI offers unprecedented opportunities, but without proper regulation, oversight, and societal debate, the balance between progress and risks could dangerously shift out of focus.
Comments