dramaWTF 7.3via Hacker News Frontpage
Blatant AI slop just won a 25k USD DeepMind Kaggle Grand Prize
"Winning $25k for the ultimate 'trust me bro' prompt is the new benchmark for AGI testing."
Explain Like I'm Normal
A winning entry in a Kaggle competition for measuring AGI progress was allegedly found to be nothing more than hard-coded 'slop' rather than a generalizeable model. The submission bypassed actual problem-solving by using specific prompt-engineering tricks to match the hidden test cases, sparking an uproar over leaderboard integrity and prize payout criteria. This incident highlights how easily current automated benchmarks can be gamed through brute-force simulation rather than intelligence.
#kaggle#deepmind#prompt-injection#gaming-the-system#slop
GET THE DAILY CHAOS
The only newsletter for people who read AI news at 3am and feel things. One email a day.