ISTANBUL
In the wake of artificial intelligence chatbot ChatGPT’s headline-grabbing release just last year, AI advocates and naysayers alike agreed on one thing: We are in the middle of rapid, revolutionary change.
In the span of a year, AI platforms have made vast progress in several areas, showing the speed at which the sector operates, despite various long-standing fears from the risk it poses to jobs to its alleged existential threat to humanity.
Will AI also come to dominate our lives as a whole? That’s a question yet to be answered, but the breakthroughs it has seen in the last 365 days attest to such a possibility.
Text trailblazer
Text has been one of the strong suits of AI chatbots, now able to process and respond to natural language prompts, carrying out a variety of tasks for users, from editing and translating to writing original responses in a matter of seconds.
These fields are expected to be dominated by AI in the near future, with the work of highly sophisticated ChatGPT-4, the full version of the popular AI tool, nearly indistinguishable from most human writers.
AI can even be asked to write in a particular style and tone, such as that of playwright William Shakespeare, who could, for example, be made to describe this phenomenon — and the dangers it may entail — thusly:
“Verily, AI doth emulate the quill of mortal wordsmiths, merging man and machine in harmonious union ... as the realm of technology unfolds, their prowess shall ascend, imbued with wisdom and innovations yet unknown.”
As “ShakespeareGPT” said, the possibilities are boundless, but it has been known to carry risks as well.
Students have been using the tool to auto-generate papers and news editors to write reports optimized for online visibility, resulting in rising concerns that integrity may suffer.
In one instance, an unnamed professor in Texas said he failed his entire class after asking ChatGPT whether his students’ papers were its work. The platform falsely claimed that it was the author of every single essay.
More recently, a Turkish daily had ChapGPT write an article but forgot to remove a disclaimer by the chatbot saying it only had information dating back to 2021.
Considering that AI is still in its early days, it is probable that solutions will be found for these blunders. Even today, it is clear that the right amount of proper human-machine collaboration proves useful for anyone dealing with text.
Another related area that AI-generated text has shown promise is coding.
DeepMind’s AlphaCode, released in 2022, has been able to beat over 70% of human coders in average competitions.
The same company also released Gato last year. The deep neural network can be used for a range of complex tasks, from playing video games to engaging in conversations, controlling a robot arm, stacking blocks, and more.
Visual virtuoso
Besides chatbots that respond to text prompts in kind, AI platforms can now also be used to produce images of exceptional quality, such as a photograph made by German artist Boris Eldagsen that won at the Sony World Photography Awards this year.
Major platforms include DALL-E, one of the most sophisticated AI image generators, and its successor DALL-E 2, released in mid-2022. Both of these were developed by OpenAI, the company behind ChatGPT.
They have been used to generate a wide range of images, from extended versions of timeless classics to brand new — and often nonsensical — illustrations, such as of teddy bears working on the moon.
Another well-known platform that generates digital images using natural language descriptions is Midjourney, which has become increasingly popular among tech enthusiasts.
It made headlines last year after a Colorado designer used it to make an image that went on to win an annual fine art competition in the state, beating human-made competition.
Medical marvel
Despite fears that AI could herald doomsday for the human race, it appears today that it is more of an asset to our well-being than a risk.
AI tools like NYUTron have proven crucial for patients in hospitals affiliated with New York University.
Developed by a team at the university’s medical school, it is able to “learn” from doctors’ notes without needing specially formatted data to accurately estimate patients’ risk of death, the duration of their hospitalization, and other factors important to their treatment.
Meanwhile, Google and Mayo Clinic have also announced a partnership to bring generative AI technologies to medicine.
Their initiative aims to allow staff to interpret important data such as a patient’s medical history, imaging records, genomics or lab results more quickly and just using a simple query, even if the information is stored across different formats and locations.
AI has also played direct roles in patient treatment.
Neuroscientists and neurosurgeons in Switzerland were able to reestablish the connection between the brain and spinal cord of a patient paralyzed in a motorcycle accident.
By surgically inserting electronic implants in movement-oriented parts of the patient’s brain and spinal cord, they were able to bypass damaged tissue, allowing him to once again move his limbs independently with the help of AI reading and transferring signals.
“Thanks to algorithms based on adaptive artificial intelligence methods, movement intentions are decoded in real time from brain recordings,” said Guillaume Charvet, one of the researchers, in a statement on their success.
Legal lifesaver
With ethics again a major concern, AI has made its way into the legal area, too.
Dozens of AI apps exist that are used in legal practices, including Latch, which uses OpenAI’s ChatGPT-4 to modify and make amendments to negotiation agreements.
The Microsoft Word-based add-in can generate suggestions, modify clauses, and answer legal questions based on a given contract.
Others have been created to sift through legal documents for useful information, draft new ones, automate repetitive tasks, and enter data.
Yet AI still falls short of serving more complex functions in the legal sector.
One New York lawyer made the news recently after he failed to check fake legal cases that ChatGPT made up when he asked it for help writing up an affidavit.
The mistake became apparent in court, when the lawyer was grilled by the judge over his oversight.
These mishaps and others like them show the extent of expectations for AI, even at this early stage.
As work continues to further develop these programs and technology advances in the coming years, platforms are poised to push their boundaries even more and bring the future ever closer.
