NVIDIA today signed a memorandum of understanding (MoU) with a consortium of universities in Thailand to drive research and accelerate scientific breakthroughs in artificial intelligence (AI) and high performance computing (HPC). The national collaboration is driven by Thailand’s aim to create a [...]
From modifying food to making fake news, AI is changing the world in different ways than one would think.
Glossy prototypes of driverless cars and Sophia the Robot continue to dazzle the world. Against this backdrop, artificial intelligence (AI) has been steadily growing its roots within a tech-enabled civilization.
Companies are also using AI to solve structural world problems. Rainforest Connection, for instance, uses AI-enabled hardware perched atop trees to detect and prevent illegal logging in different areas of the world.
Inventors have stretched the reach of companies working with AI, putting financial heft behind solving complex problems, and encouraging innovators to set new challenges for themselves. Some of these creative applications of AI may take you by surprise.
With AI, you can now have your cake, eat it too, and feel none of the guilt in the aftermath. AI continues to have an increasing impact on foodtech and agritech, driving trends in these industries. One of the more unique ways in which this is happening is through food reformulation.
Swiss flavors and fragrances developer Firmenich recently announced that it had developed the world’s first AI flavor. The company described the flavor as “a delicious lightly-grilled beef taste for use in plant-based meat alternatives.”
Another foodtech company, Singapore-based Hoow Foods, uses a machine learning-based product development platform to analyze food products. The platform can even modify product formulations to change taste, texture, and nutritional value, the company’s website states.
The company’s technology works on ingredients such as sauces, condiments, and staples, as well as finished products such as ice-cream.
Technology that can customize food to this molecular level is powerful in many ways. While it still has a long way to go before it’s adopted by end-users, it marks a turning point in the way food is grown, the relationship between food and chronic conditions such as diabetes, and global food consumption patterns.
These tiny, yellow-black flying insects have become the subject of much public attention over the years due to their declining numbers.
Declines in the bee population in recent years has far exceeded the average 5%-10% that hives tend to lose in winters. Owing to their crucial role as pollinators, this has further upended the world’s ecological balance.
The problem is still reversible, but researchers have already developed another solution for a doomsday scenario – robot bees.
Essentially, these are tiny drones that perform the vital pollination function bees execute in the environment. (Fun fact: male honeybees are also called drones!)
Through a combination of micro robotics, aerodynamic design, adhesives, and AI, robot bees should be able to replicate pollination, or so theory suggests. Yet, bee function is too complex for current robo-bee technology to perfectly mimic.
While several projects such as this one at Harvard University are underway, robot bee technology needs to address several shortcomings, including costs, unsustainable models, and the question of ecological balance, before deploying artificial bees at scale.
Moreover, robot bees are a fascinating scientific invention, but perhaps the world’s scientific efforts will be better employed in protecting the real ones.
Generating fake news
Fake news stories are weapons of mass disinformation. With natural language generation (NLG), AI can make it possible for anyone to generate fake news with just a few clicks.
NLG uses AI and deep learning algorithms to convert data into simple, plain human language. It can be applied to a variety of scenarios, such as writing songs, responding to customer queries, or more sinisterly, writing deliberately inaccurate news.
Against a backdrop of global political uncertainties, false information – whether engineered or shared unintentionally – poses a significant threat to political stability and public safety.
Sites such as OpenAI’s GPT-2 or Grover are text generation AI models that train on articles that already exist on the web. They learn from these models to string words together words that sometimes don’t make sense, but are fake nonetheless.
The more these models learn, the better they get at creating coherent (yet false) articles. Pair this with the fact that AI can also learn to reflect human bias based on training datasets, and the recipe for an infodemic disaster is complete.
Judging a beauty contest
Backed by partners such as Nvidia and EY, Beauty.AI launched the world’s first global beauty pageant judged by AI in 2015.
Here’s how the eccentric competition works: users download the app and upload a selfie without make-up, glasses, or a beard. The competition’s robot jury then evaluates the photos and crowns a beauty Queen and King.
The jury was composed through an application process inviting data scientists to submit their algorithms as members of the jury. Selected algorithms were trained to analyze submitted photos on parameters that include wrinkles, face symmetry, skin color, gender, age group, and ethnicity.
When the second leg of the competition, Beauty.AI 2.0, launched in 2016, around 6000 photos were submitted from over 100 countries. The results were immediately called out for being starkly racist – the list of 44 winners was predominantly white, featuring only a few people of color.
Beauty.AI Chief Science Officer Alex Zhavoronkov attributed this to a lack of minority data in the datasets used to train the AI jury.
Youth Laboratories, the creators of the contest, announced that Beauty.AI 3.0 was in the making. However, a spokesperson tells Jumpstart that Youth Laboratories has postponed Beauty.AI 3.0 due to a change in plans.
The creators of Beauty.AI have now shifted gears to a skin SaaS platform. Called Haut.AI, the platform “uses AI to analyze skin health and aging biomarkers and provide skincare and lifestyle recommendations,” Anastasia Georgievskaya, General Manager of Beauty.AI and CEO of Haut.AI, said in an email exchange.
From finding music to making it
British artist scouting app Instrumental uses AI to collect and check data on emerging indie artists. The company’s algorithms help it find the fastest growing unsigned artists online. Instrumental invites these independent artists to join the Instrumental network.
Instrumental’s data science capabilities also extend to real-time data on streaming, social media, audience demographics, and fan engagement. It goes a step further by matching indie artists with commercial partners.
Instrumental has worked to connect prominent labels including including Live Nation and Sony Music with independent artists. Further, the company will reportedly be welcoming Chinese gaming giant and corporate juggernaut Tencent Holdings as a minority stakeholder in the company.
However, not only can AI help to identify high-potential music creators, it can create music too.
Alan Turing recorded the first computer-generated music way back in 1951, and this application of AI continues to exist. Today, AI uses data from different compositions to decipher which elements are the most enjoyable. It can also mimic a genre or compose songs by combining different elements of music.
Several platforms, such as Amper Music, AIVA or MuseNet (created by OpenAI, the same developers behind the fake news generator GPT-2), make this composition technology accessible over the Internet, for a price or even for free.
Use cases for artificial intelligence have expanded to cover every facet of daily life on a deep level. This is slowly creating a new paradigm, where AI does the routine heavy lifting so that human intelligence can be reserved for more fulfilling tasks.
As this shift takes place, it’s important to remember that these granular applications of AI do not create something out of nothing. They echo the human intelligence within the datasets that the AI learns from. In doing so, they serve as a reminder that the student can indeed become the master.
It is important, thus, for the masters to get their lessons straight.
Header image by Kaboompics .com from Pexels