Semantics and other AI problems we can solve
Chatbots, speakers, translation machines… Human-like artificial intelligence is nothing new to our world. In fact, the machine’s actions often come so close to human reality that one could wonder: where does its capabilities end? Is there anything that AI cannot do? Probably not. But there are thresholds on the road that we can only bridge using the strongest of resources. One of those thresholds is semantics.
Literally, we define semantics as “the study of meaning”, or the science that studies the meaning of words. This goes deeper than just what is written in the dictionary - especially if you look at a word in its context within a written or spoken text. Take the word ‘bank’ for instance. If you say you live along the banks, we know you live near a river and not on a street filled with financial institutions. But can an AI also understand that? It’s not that easy. Because deep learning AI may excell very well in maths, it does not necessarily have a knack for language.
More than a definition
On top of all that, the meaning of what you say is not only hidden in its definition, but also in how you bring it. Does your sentence carry a certain emotion that is relevant to its meaning? For us, humans, it seems logical that we understand this, but it actually makes no sense. Take a look a this example:
Because of that, meaning is not something a machine can easily extract from data, making Natural Language Understanding one of the hardest but also most interesting domains of AI. For years, it has been a challenge for (small and big) tech companies to cross the threshold that ‘meaning’ proves to be. A threshold that even Open AI’s GPT-4 and IBM’s Watson can’t seem to bridge without failing.
Back to basics
Some scientists argue that the AI-gap to understand meaning is not in code or data, but in the prelinguistic basics that we are born with as humans. Primal principles that are woven into our instincts and give us a sense of space, time and other essential concepts that allow us to understand certain connections - with or without language. Connections that may seem logical to us, but are not at all logical to machines.
So, should we forget everything we have built up so far and take AI back to basics? Return it to a kind of infantile state in which we can teach it that solid semantic basis? Perhaps, according to some, who then speak of "infant metaphysics".
“Training and evaluating machines for baby-level intelligence may seem like a giant step backward compared to the prodigious feats of AI systems like Watson and GPT-3. But if true and trustworthy understanding is the goal, this may be the only path to machines that can genuinely comprehend what “it” refers to in a sentence, and everything else that understanding “it” entails.”
Semantics and AI: practical uses
At Nalantis, we have grabbed the challenges posed by Natural Language Understanding with both hands. For over ten years now, we have been continuously improving our NLU applications, including semantic analysis of written and spoken text.
As a result, we can now say that our next-gen language technology is used to convert text into meaning so that machines can understand it. Thanks to our No Black Box AI approach, we can also do this transparently and with constant improvements to the process.
It may sound abstract, but our applications are certainly not. For example, our technology is being used to convert recordings of municipal council meetings into data containing the decisions that have been made. This data can then be consulted by employees and citizens who have questions about certain municipal decisions. An open government approach? Yes, please. And let's make it easier while we're at it.
Another example: our role in the Flying Forward 2020 consortium. The challenge here is to make legal texts (that are written by humans) understandable to machines, such as autonomous drones. If we succeed, this will mean a lot for the future of Urban Air Mobility and will make it easier to put urban air transport on the European map.
Curious about what the future holds for AI and semantics? Stay in touch with us.
