Lifeboat Foundation

Meta has created an AI language model that (in a refreshing change of pace) isn’t a ChatGPT clone. The company’s Massively Multilingual Speech (MMS) project can recognize over 4,000 spoken languages and produce speech (text-to-speech) in over 1,100. Like most of its other publicly announced AI projects, Meta is open-sourcing MMS today to help preserve language diversity and encourage researchers to build on its foundation. “Today, we are publicly sharing our models and code so that others in the research community can build upon our work,” the company wrote. “Through this work, we hope to make a small contribution to preserve the incredible language diversity of the world.”

Speech recognition and text-to-speech models typically require training on thousands of hours of audio with accompanying transcription labels. (Labels are crucial to machine learning, allowing the algorithms to correctly categorize and “understand” the data.) But for languages that aren’t widely used in industrialized nations — many of which are in danger of disappearing in the coming decades — “this data simply does not exist,” as Meta puts it.

Meta used an unconventional approach to collecting audio data: tapping into audio recordings of translated religious texts. “We turned to religious texts, such as the Bible, that have been translated in many different languages and whose translations have been widely studied for text-based language translation research,” the company said. “These translations have publicly available audio recordings of people reading these texts in different languages.” Incorporating the unlabeled recordings of the Bible and similar texts, Meta’s researchers increased the model’s available languages to over 4,000.

On Wednesday, Google displayed how Bard, its new AI robot, could be used to write up job listings from a simple one line prompt. Microsoft has demonstrated how a ChatGPT-powered tool can write entire articles in Word.

“There are a tonne of sales representatives doing a lot of banal work to compose prospecting emails,” says Rob Seaman, a senior vice president at workplace messaging company Slack, which is working with OpenAI to embed ChatGPT into its app as a kind of digital co-worker.

New AI tools may remove some of the most tedious aspects of such roles. But based on past evidence, technology also threatens to create a whole new class of menial roles.

Perspective from a very-educated layman. Er, laywoman.

This is Hello, Computer, a series of interviews carried out in 2023 at a time when artificial intelligence appears to be going everywhere, all at once.

Continue reading “Hello, Computer — Sabine Hossenfelder — A.I. going mainstream” »

The other day a friend proudly told me she wrote a heartwarming graduation card to her teenage son. “Okay,” she confessed. “I

How long was your card? I asked her.

Not only that, but many also couldn’t even generate a topic on their own. They lacked creativity to dream up their own ideas, much less the critical thinking skills to put themselves in the shoes of their audience, imagining what would land. But they all had 4.0 GPAs or higher and came from private schools in Orange County and LA, reflecting our watered-down educational system.

Continue reading “Will AI Produce Useless Humans?” »

With LIMA, Meta’s AI researchers introduce a new language model that achieves GPT-4 and Bard level performance in test scenarios, albeit fine-tuned with relatively few examples.

LIMA stands for “Less is More for Alignment,” and the name hints at the model’s function: It is intended to show that with an extensively pre-trained AI model, a few examples are sufficient to achieve high-quality results.

Few examples in this case means that Meta manually selected 1,000 diverse prompts and their output from sources such as other research papers, WikiHow, StackExchange, and Reddit.

Summary: Researchers used artificial intelligence to predict how enzymes interact with various substrates. The team developed an AI model that can accurately predict whether an enzyme can work with a particular molecule.

Their enzyme substrate prediction (ESP) model provides a valuable tool for drug research and biotechnology, with applications ranging from the creation of new drugs to the production of biofuels.

I asked Bard chatbot whether we should fear Google. It shared concerns about Google you’d never expect would come from an AI created by Google.

Plus: Official ChatGPT iPhone app debuts; Debt collectors using chatbots to chase debtors.

ChatGPT has changed the world since it emerged a few short months ago. Where will future advancements in generative AI take us?

Welcome back Katie Brenneman, a regular contributor to 21st Century Tech Blog. Several weeks ago when ChatGPT entered the headlines I suggested to Katie that she consider writing about Large Language Modelling (LLM) and the technological and societal implications in terms of its capabilities. Were we witnessing the birth of consciousness in this new artificial intelligence (AI) discipline, or were we coming to terms with what defines our sentience?

By definition, sentience is about feelings and sensations and not thinking. Consciousness, on the other hand, is about our awareness of self and our place in the world around us. And thinking is about the ability to reason, consider a problem, come up with an idea or solution, or have an opinion.

Continue reading “Are Large Language Model Generative AIs Sentient, Conscious or Thinking?” »