Artificial intelligence (AI) has permeated every sphere of our civilization and way of life over the past ten years.
It’s difficult to deny its influence on everything from chatbots and virtual assistants like Siri and Alexa to automated industrial gear and self-driving automobiles.
Machine learning consists of sophisticated software algorithms designed to perform a single specific task, such as answering questions, translating languages, or navigating a journey.
And get better at it as they are exposed to more and more data, which is currently the technology most frequently used to achieve AI.
According to IDC research, governments and businesses will spend more than $500 billion on AI technology globally in 2023.
But what effect will it have, and how will it be used?
Here, we summarize the critical societal and corporate themes that will shape the usage of artificial intelligence over the upcoming 12 months.
Next 10 years AI prediction by Andrew Ng:
This article from March 21 by George Anandiotis features an interview with Andrew Ng, co-chairman and co-founder of Coursera, creator of Landing AI and DeepLearning AI, and adjunct professor at Stanford University.
Big data has received much attention in the past ten years, and Ng told VentureBeat. He anticipates a trend toward data-centric AI in the following decades.
“I underestimated the amount of work that would be needed to flesh out deep learning 10 years ago, and I think many people are underestimating the amount of work that would be needed to flesh out data-centric AI to its full potential today,” Ng added.
“But it will enable many more AI applications as we jointly advance on this over the next several years, and I’m incredibly thrilled about that.”
Meta layoffs expected infrastructure:
On November 9, the day Meta announced it was laying off 1,000 staff, senior writer Sharon Goldman was up late at night going through Twitter.
Mark Zuckerberg shared a message with Meta workers in a public announcement, which some took to mean that individuals employed in artificial intelligence (AI) and machine learning (ML) may escape the worst of the layoffs.
However, Thomas Ahle, a Meta research scientist who was fired, tweeted that he and a research group that concentrated on implementing machine learning throughout the infrastructure stack, had been eliminated.
He stated that the team consisted of 50 people, excluding the managers.
OpenAI’s GPT-4 Rumors:
A new model in the GPT-3 family of AI-powered large language models (LLMs), ChatGPT, outperforms its predecessors by processing more complicated instructions and delivering higher-quality, longer-form material.
At the same time, GPT-4 rumors persisted at NeurIPS 2022 on November 30.
Even though ChatGPT has only been available for a short while, it has continued creating headlines.
DeepMind fast Matrix Multiplication Algorithms:
One of the essential jobs in machine learning, matrix multiplication, was one of the most complex mathematical riddles to solve.
Could AI develop its methods to speed up matrix multiplication?
Research company DeepMind revealed AlphaTensor, the “first artificial intelligence system for developing unique, efficient and provably right algorithms,” in an article in Nature on October 5.
The Google-owned lab said that the study “sheds light” on a mathematical problem that has been unsolved for 50 years: how to multiply two matrices as quickly as possible.
A DeepMind blog article claims that AlphaTensor is based on AlphaZero, an agent that has demonstrated superhuman performance in board games like Go and Chess.
The AlphaZero adventure is furthered in this new work, which switches from playing games to solving unsolvable mathematical puzzles.
Google’s Simple ML for Sheets:
Sean Michael Kerner announced Google’s intention to integrate machine learning into its Sheets tool on December 7.
Machine learning (ML) has yet to be integrated into spreadsheet experiences as simple calculations and graphs have.
Spreadsheets are designed to be usable by any user, whereas ML is frequently perceived as too hard to use.
According to Google, the Simple ML for Sheets add-on has received a beta release.
Due to the extensible nature of Google Sheets, users may use add-ons that enhance the program’s basic features.
Under this instance, Google Sheets gains from machine learning (ML) technology that Google first created in the open-source TensorFlow project.
Users won’t need to use a specialized TensorFlow service with Simple ML for Sheets because Google designed the service to be as user-friendly as feasible.
Deep learning Revolution:
Senior writer Sharon Goldman contacted AI pioneer Geoffrey Hinton when she noticed that September 2022 was the tenth anniversary of necessary neural network research, known as AlexNet, that sparked the profound learning revolution in 2012.
Hinton and other prominent AI figures, such as Yann LeCun and Fei-Fei Li, were interviewed for this article, which also takes a close look at the future of AI.
OpenAI’s DALL-E 2:
The strong image-generating AI system DALL-E 2’s increased beta access by OpenAI in late July excited the tech community but also raised numerous doubts.
What, for instance, does the commercial exploitation of DALL-AI-powered E’s photography imply for the creative workforce and industries? Will it take their place?
The response, according to OpenAI, is no. According to an OpenAI official who talked with VentureBeat, DALL-E is a tool that “enhances and extends the creative process.” DALL-E can aid an artist in coming up with original ideas, much like how an artist would look at other art pieces for inspiration.
Regarding the ownership of photos produced by AI, discussion and criticism have persisted after this article was published. There is no doubt that it won’t end soon.
Intel unveils real-time deepfake detector:
The first real-time detector of deepfakes, or synthetic media in which a person in an existing image or video is substituted with someone else’s likeness, was unveiled by Intel on November 16.
The gadget, according to Intel, operates by monitoring the minute “blood flow” of video pixels to produce findings in milliseconds and boasts a 96% accuracy rate.
This kind of deepfake detection technique is becoming increasingly crucial as deepfake dangers loom.
Who owns DALL-E images:
Senior writer Sharon Goldman investigated the legal implications of devices like DALL-E 2 in another episode of what has become an ongoing text-to-image generating drama.
The business granted paying subscription customers exclusive use rights to reproduce, sell, and market the pictures they produced with the potent text-to-image generator when OpenAI announced expanded beta access to DALL-E in July.
After a week, creative experts from various sectors started to ask plenty of questions.
First on the list: Who owns the pictures produced by DALL-E and other AI-powered text-to-image converters like Google’s Imagen, for that matter?
The AI’s creator, who also trained the model? Or the person commanding the AI?
Who owns the DALL-E photos is not immediately evident, according to Bradford Newman, who oversees Baker McKenzie’s machine learning and AI business at its Palo Alto office.
And the inevitable legal repercussions, he underlined.
This post by Sean-Michael Kerner, dated August 23, discusses the Ray architecture that underpins OpenAI.
The open-source Ray framework, utilized by businesses like OpenAI, Shopify, and Instacart, has become one of the most popular solutions for corporations to scale and manage increasingly vast and complicated artificial intelligence workloads over the past two years.
In addition to supporting MLops workflows across various ML tools, Ray makes it possible for machine learning (ML) models to scale across hardware resources.
The Ray Summit in San Francisco saw the presentation of the tool’s upcoming significant milestone.
With the addition of the new Ray AI Runtime (AIR), which serves as a runtime layer for ML service execution, Ray 2.0 expands the technology.