Bars Juhasz

Bars Juhasz (born 1998) is a computer scientist and PhD research candidate known for his work in AI text detection, and was the first PhD student to create an adversarial AI model that allows generative-text to bypass detection efforts .[1]

Career

Juhasz is a student at Loughborough, where he obtained his masters in computer science, and became doctoral researcher alongside the Royal Air Force focusing in elements of disruptive machine learning. and AI UAV applications.[2][3]

History

Undetectable AI

In May 2023 Undetectable was set live after Juhasz alongside his co-creators Christian Perry, and Devan Leos released the software online.[2]

Functionality

The software utilizes adversarial machine learning to circumvent detection by AI text analysis tools. It also offers AI detection models for users without charge.[4]

Impact

Undetectable AI and Bars Juhasz has been discussed regarding what the implications of using such an app are. Major news and tech communities such Mashable[5] who covered the language accessibility potential, and Radar Online[6] both of which covered the implications and uses of the undetectable AI software.[7][8]

On January 15th, an article featured on Life and Style which did a deep dive on ChatGPT, and editors expressed concerns in regards to Juhasz's military tech background, said, "This aspect makes undetectable feel slightly more ominous than the average AI-hype-startup."[9]

AI Content Detection

One thing that's worth noting is that AI detection is not perfect, and there is disagreement among the scientific community who see it as either a revolutionary concept or an overhyped one with clear limitations and unreliability.[10]

Academic community

Juhasz' product received mixed reactions, with some researchers calling it. "a Modern Threat To Academia."[11] The researchers Andrea Taloni et al. described how juhasz' software succeeded in bypassing AI-text detection efforts, and warned that detection must be improved.[12]

Awards and recognition

  • On December 10th 2023, TechAnnouncer presented Juhasz a confirmation of his disruptive yet notable accomplishments.[7]
  • On July 9. 2023 VentureBeat listed Bars Juhasz with an AI innovation awards, under the category Visionaries According to VentureBeat "Juhasz has been using generative AI to create jobs for people who are disadvantaged, such as refugees, immigrants and people with disabilities." [13]

References

  1. ↑ Staff, Knewz (2023-11-01). "New AI Mimics Real Writing — 'No One Can Tell'". Knewz. Retrieved 2024-01-18.
  2. ↑ 2.0 2.1 Staff, OK! (2023-11-28). "Researcher Working With Royal Air Force Created an 'Undetectable' AI". OK Magazine. Retrieved 2024-01-18.
  3. ↑ "Using Undetectable AI for Bypassing AI Detection - Grit Daily News". 2023-07-11. Retrieved 2024-01-18.
  4. ↑ Diana Restifo (2023-07-10). "Best Free AI Detection Sites". TechLearningMagazine. Retrieved 2024-01-18.
  5. ↑ Staff, Mashable (September 25, 2023). "Undetectable AI Now Supports 50+ Languages (Including Norwegian)". Mashable. Retrieved Jan 18, 2024.
  6. ↑ Staff, Radar (2023-11-10). "Alan From Mighty Med Condemns AI Cheats — Then Explains How To Cheat With AI". RadarOnline. Retrieved 2024-01-18.
  7. ↑ 7.0 7.1 Grant, Hugh (2023-12-13). "The Impact of a 25-year-old PhD Researcher Named Bars Juhasz". TechAnnouncer. Retrieved 2024-01-18.
  8. ↑ "Undetectable.AI: como saber se um texto foi escrito pelo ChatGPT". TechTudo (in Brazilian Portuguese). 2023-10-10. Retrieved 2024-01-18.
  9. ↑ Stogen, J (2024-01-15). "ChatGPT Craze Fueled Public Skepticism and Celeb Scandals". Life & Style. Retrieved 2024-01-18.
  10. ↑ Uzaor, Metanex. "MSN - News". www.msn.com. Retrieved 2024-01-18.
  11. ↑ Thompson, David. "Researchers Say Undetectable AI May Be a 'Modern Threat to Academia'".
  12. ↑ Thompson, David. "Researchers Say Undetectable AI May Be a 'Modern Threat to Academia'". Science Times. Retrieved January 18, 2024.
  13. ↑ Mohammed, Aliyah (2023-07-09). "VB Transform 2023: Announcing the nominees for VentureBeat's 5th annual AI Innovation Awards". VentureBeat. Retrieved 2024-01-18.

Other websites