Grasping the sociological import of AI, as it exists today, and what it incubates is a spectacle. AI doesn’t materialize inside some remote or far-flung sci-fi future: it originates now, within the social matrix we inhabit, regardless of our habitus. And with every effort and experiment, we are training the models deeper and unbeknownst to every user.
In a world already grappling with the twin transitions of digitalization and decarbonisation, a double illusion is at play; a misapprehension that represents AI as cognition and intelligence.
But AI is not awareness, not synaptic linkages. AI is a deep entanglement of social relations. The operational concepts surrounding AI are optimisation, layers, vectors, and loss functions that measure the difference between current and desired output. And although many benefits may not materialize immediately, investors are eager to bet on an AI-driven future.
Outsized gains have come with the expectation of huge earnings for a handful of massive growth and technology names that have dominated the US stock exchange. Dubbed “The Magnificent Seven”, the shares of Apple, Microsoft, Alphabet, Amazon, Nvidia, Tesla, and Meta have soared between 40% and over 200% during 2023.
Despite the global economic slowdown resulting from Sars-Cov-2, the seven Nasdaq behemoths have thrived. We are at the edge of a dive to understand generative AI’s power, reach, and capabilities. AI has altered the anatomy of creative work, communication, digital trade, and study.
Generative AI applications like “GitHub Copilot”, “Stable Diffusion”, and “ChatGPT”, have unlocked the imagination of students, workers, and creatives around the world. Thanks to their broad utility, almost anyone in rural or urban settings can use these applications to create and communicate.
These applications can perform a range of routine tasks, but it is their broad utility that has persuaded consumers and households to experiment with them on their own at school and at work. The large language model (LLM) called GPT-4 has entangled the hearts and hands of humans into a collective.
“GitHub Copilot” is an AI pair programmer. It offers auto-complete style suggestions as you write code. Prompts from “GitHub Copilot” are received either by starting to write the code you desire or by writing a natural language comment describing what you wish the code to accomplish.
Using text input, “Stable Diffusion” can generate photo-realistic pictures. It is a deep-learning text-to-image diffusion model. In May 2023, Google announced several new features including “Search Generative Experience” using generative AI, and a new large language model (LLM) called “PaLM 2” that powers “Bard” and other Google innovations.
This new wave of AI applications has left Siri, Cortana, and Alexa behind. The new generative AI applications can process large and varied sets of unstructured data, and perform more than one task across a range of modalities, including images, audio, and code. AI trained on these new foundation models can classify, edit, summarize, answer questions, and draft fresh content.
New AI Assistants – frequently called “agents” or “copilots” perform complex tasks like creating investment strategies without close human supervision, renting a flat on Airbnb, and summarizing meetings for those who were delayed in other consultations.
These “copilots” can deliver a Microsoft Teams engagement knowing the preferences of the participants, resolving calendar conflicts of attendees, and rescheduling actions using priorities while maintaining the careful touch of client sensitivities.
“Adept”, an AI startup, in an online demo, shows how users can prompt its technology with a sentence and then observe how it navigates a company’s salesforce customer-relationship database on its own. “Inflection AI”, a one-year-old startup that released the chatbot “Pi” in May 2023, just raised US $1.3B from investors like Nvidia and Microsoft in a mix of cloud credit and cash.
Founded by Google DeepMind co-founder Mustapha Suleyman, and LinkedIn cofounder Reid Hoffman, “Pi” is focused on consumer-facing AI products. “Pi” is now the uppermost rival of OpenAI. Unafraid of the future, a report on Inflection-1, the model that powers “Pi” was released recently. It claims that “Pi” has outperformed the competition.
The venture capital which “Pi” has attracted will be used to build computing power for a more powerful foundation model. At the 2023 Collision Conference, Mustapha Suleyman told participants that the intention is to build a cluster of about 22,000 H100s. This will be three times more computing power than was used to train all of GPT4.
Generative AI’s impact on productivity could contribute trillions of dollars in value to the Latin American and the Caribbean economy. Generative AI can be used to support customer interactions, digital trade, generate creative content for marketing and sales, and assist with computer code based on natural-language prompts to advance a CARICOM Cloud initiative.
Mobile payments, e-Commerce, automated investment, and portfolio management services are among the activities that could see the biggest impact as a percentage of their revenues from generative AI. In tourism, the digital attractiveness of destinations can now be amplified with generative AI.
Early pilots are compelling. But a complete understanding of the technology’s benefits will take time. Society will have considerable challenges to address, including managing the workforce risks inherent in generative AI and reimagining education for a generative AI curriculum in a deschooled society.
