Tooploox CS and AI news 24

  • Scope:
  • AI
  • Artificial Intelligence
  • Artificial Intelligence|Business
  • Artificial Intelligence|Business|Technology
  • Artificial Intelligence|Technologies in IT|Technology
  • Artificial Intelligence|Technology
Tooploox CS and AI news 24
Date: December 15, 2022 Author: Konrad Budek 3 min read

The last days of November kickstarted another month with NeurIPS – arguably the largest and the most significant AI-related conference in the known universe. 

Apart from the conference, this edition of the news digest also comes with info on analog neural networks (not to be confused with the biological variety) and a real look at the way robots take over human jobs.

Tooploox on NeurIPS 2022

The Conference on Neural Information Processing (NeurIPS) is one of the most important AI and ML-related events in the world. This year’s conference is being held in New Orleans, Louisiana for the first week and over the internet, delivered via digital component, for the second week. 

The conference started on Monday, November 28 and ends on December 9. As with many years past, Tooploox continues to support global endeavors toward building better neural networks and improving machine learning-based solutions by delivering research papers at the conference. This year, the team has delivered a way to increase the accuracy of predictions delivered by Hidden Markov Models. Detailed information about Tooploox’s works and research can be found in our blog post about Improving Hidden Markov Models.

The conference itself has been held for 36 years and keeps the interdisciplinary approach of gathering researchers from various fields, united by the common goal of building a better understanding of neural networks. Currently, the lineup of the conference is dominated by machine learning and artificial neural network researchers. 

More information about the conference can be found on its website.

Overcoming the limitations of analog neural networks 

With the renaissance of vinyl records and the understandable nostalgia for cassette players, building analog neural networks sounds unsurprising – yet this does not even approach the reasons why. These networks are based on optical signals instead of digital ones, making the whole system significantly faster. 

The challenges are in scaling up the structure, as these structures are prone to hardware errors that tend to stack up with scale. At the core of the structure, there are Mach-Zehnder Interferometers (MZI) which can be compared to tunable mirrors used to channel and process light signals. These are the primary sources of hardware errors. 

Researchers from MIT improved the chip construction to help it not lose the power of the beam and thus significantly reduce the number of errors occurring in the process. More about the technology can be found in the MIT blog post. 

Employees willing to delegate mundane tasks to AI

According to a recent study done by Harris Poll, up to 72% of employees are willing to delegate mundane, dull, and repetitive tasks to ML-powered solutions. This interest is fueled not only by the dullness of these tasks but also by the fact that it can provide employees with more free time to spend with friends and family. 

This approach is also fueled by a generational shift in companies, where Generation Z is slowly infiltrating the labor market. 

The full study can be found in this press release.

Intel releases FakeCatcher 

DeepFakes are videos that have been modified to include a person who was not truly participating. The technology is widely used in the entertainment industry, for example, to place a long dead actor in a prequel movie. 

Peter Cushing, the actor playing Great Moff Tarkin in the video above, died in 1994, while the Rogue One movie was filmed in 2016. 

On the other hand, deepfakes pose a great threat when used to enhance and augment the spread of fake news. To tackle this challenge, Intel has delivered an AI-based deepfake detector. The company claims that it can spot deepfakes with 96% accuracy. 

More about the FakeCatcher can be found on their company website

Robots are taking over jobs

Researchers from Brigham Young University decided to take a closer look at the ways robots actually replace human workers. According to their work, up to 14% of employees have already seen their job positions being taken over by robots. 

Among those who have seen their job position replaced by robots, the perceived effect is overstated by a factor of three. From their viewpoint, up to 47% of all existing jobs will be filled by robots. 

The research says that this fear is exaggerated and is obscuring the real way robots will work alongside human workers – the fact being that machines actually augment the workforce rather than directly replace human talent. 
More about the research can be found in ScienceDaily.

Latest posts

See all posts