Witz (4440) · RiskRestated
Incidents Arising from Opaque AI Decision-Making and the Difficulty of Demonstrating Safety
First written in FY2019. 2 of 3 entries say "Restated". Latest entry: FY2025.
3entries
7years
0from disclosures
How it moved
Same theme, other companies
Risk of Malfunctions and Deficiencies in New Technologies Such as Cloud Computing, Blockchain and AIA theme 6 companies write about · 6 as a headwind · 0 as a tailwind
3 other companies' latest entry also says "Restated".
Ridges = years it wrote progress, valleys = years it wrote worse. Most hopeful companies at the back. Drag to turn, scroll to zoom, right-drag to pan. On a phone, two fingers turn and pinch to zoom; one finger scrolls the page.
Full recordFilings, disclosures and news, newest first
- FY2025RestatedAnnual report
「AIの急速な発展と活用により、今後増々未知のリスクが増大することが予想されます。」
AIの急速な発展・活用に伴う未知のリスクを認識し、AIセーフティ及びHMCESによる検証環境の事業化準備を進めている。 - FY2020RestatedAnnual report
「人工知能は判断結果を導き出す方法が明確でなく、一般的なソフトウェアなどに求められる安全分析は実施できません（安全分析はソフトウェアの構造が明確であり、明確な構造を順次破壊した場合の挙動を分析します。人工知能は内部構造が不明な巨大なブラックボックスであるため、従来型の内部構造が明確であることを前提としている安全分析では解析できません）。そのため従来型の分析手法を利用した方法での人工知能の安全立証は困難であります。」
人工知能の判断の不透明性と従来型手法による安全立証の困難性を、人工知能安全活用上の課題として引き続き詳細に説明している。 - FY2019First statedAnnual report
「The use of artificial intelligence in control equipment may result in serious incidents if the AI behaves uncontrollably, making the safe use of artificial intelligence extremely important. Unfortunately, AI research currently contains relatively few examples of safety-focused research, as priority has been placed on improving functions and performance.」
The Group identifies the possibility of serious incidents caused by AI behaving unpredictably in control equipment, as well as insufficient safety-focused AI research, as risks.