Generative artificial intelligence (GenAI) is a subset of artificial intelligence which uses computers to generate new content, which is often hard to distinguish from human-generated works. This content can be text (such as this article), pictures (see accompanying image), audio, videos, but also more niche and specialised artefacts such as computer code. These GenAI models learn from available data to produce new examples.
GenAI technology is not new but has made leaps forward in recent years due to two main factors. First, there has been an exponential increase in terms of available data which are used to build these models.
Second, there have been huge strides in the development and availability of specialised hardware, mostly in the form of graphics processing units (GPUs) which are well-suited for parallel computations. The stock price for NVIDIA Corp (the main manufacturer of these GPUs) has more than doubled in the past six months.
A sub-category of GenAI are Large Language Models (LLMs). LLMs focus on language (text) modelling and, for example, may predict the next word in a phrase or generate content based on some user prompt (e.g. “what are LLMs?”). These LLMs are “trained” on a broad corpus of text from the internet, books, articles, websites, etc, allowing them to acquire general knowledge across many domains.
GenAI technology is not new but has made leaps forward in recent years due to two main factors
ChatGPT is the leading example of an LLM. ChatGPT was built using texts containing 300 billion words. Occasionally, an LLM may “hallucinate” when it generates content which is not grounded in reality (e.g. output which is factually incorrect or nonsensical).
GenAI has countless applications. Businesses use GenAI to boost customer experiences by generating naturally sounding text in customer-agent interactions (e.g. automated chat bots), summarise larger documents, running personalised marketing campaigns, or creating content for their website.
GenAI also has many applications in education, where these models can act as personal tutors. We will see an increased use of GenAI-based tools with some functioning as (smart) personal assistants. An obvious application is to augment any user interface with natural language capabilities (e.g. internet search).
Jean-Paul Ebejer is an associate professor at the Centre for Molecular Medicine and Biobanking at the University of Malta. His main scientific interests lie in artificial intelligence and bioinformatics.
Sound Bites
• A pre-clinical study using stem cells to produce progenitor photoreceptor cells ‒ light-detecting cells found in the eye ‒ and then transplanting these into experimental models of damaged retinas has resulted in significant vision recovery. This finding, by scientists at Duke-NUS Medical School, the Singapore Eye Research Institute and the Karolinska Institute in Sweden, marks a first step towards potentially restoring vision in eye diseases characterised by photoreceptor loss.
• Scientists have created a new family of polymers capable of killing bacteria without inducing antibiotic resistance ‒ a major step in the fight against superbugs like E. coli and MRSA.
https://www.sciencedaily.com/releases/2023/12/231222145359.htm
For more soundbites, listen to Radio Mocha every Saturday at 7.30pm on Radju Malta and the following Monday at 9pm on Radju Malta 2 https://www.fb.com/RadioMochaMalta/.
DID YOU KNOW?
• Word of the DAY: AI-NXIETY (neologism) – feeling anxious about how AI could affect your life.
• The word ‘dragon’ comes from Greek, where the ancient root ‘dṛk-’ meaning ‘to look’ produced the participle ‘drakṓn’ (‘looking’) and then the noun ‘drakṓn’ denoting a monster whose look could kill.
• In 2020, traffickers were caught in Italy with cocaine being smuggled in individual coffee beans.
• At a number of British airports, it is more expensive to park a car than it is to park a light aircraft.
For more trivia, see: www.um.edu.mt/think.