When you purchase through golf links on our situation , we may earn an affiliate commission . Here ’s how it works .
Google’sDeepMindcan control a robotic arm to beat mere mortals at mesa tennis , a young study reports . ButFan Zhendong , the 2024 gold medallist for individual and team men ’s mesa tennis , can rest easy : Theartificial intelligence(AI)-powered robot could only dumbfound mediocre players , and only some of the time , according to the study , which was release Aug. 7 to the preprint databasearXivand has not been peer - reviewed .
automaton can now cook , clean and perform tumbling , but they fight to quickly answer to real - world environmental information .
" reach human - level performance in terms of accuracy , fastness and generalisation still rest a howling challenge in many domains , " the researchers wrote in the work .
Related:32 times artificial tidings bugger off it catastrophically improper
To master this limitation , the researchers combinedan industrial robot armwith a customized edition of DeepMind ’s ultrapowerful learning algorithm . DeepMinduses neuronic networks , a layered architecture that mimics how information is swear out in the human brain , to step by step learn new information . So far , it has beaten theworld ’s best Go role player , presage the structure of every protein in the body , crack decades - sometime mathematics problemsand more .
The system was develop to master specific aspects of the game — for instance , learning the rules , produce top spin , delivering forehand shot serves or using backhand shot place — training on real - earth and simulated data point in sophisticated algorithm . As the AI learned , the researchers also pull in data on its strengths , failing and limitations . Then , they fed this information back to the AI program , thus give DeepMind ’s unnamed factor a realistic impression of its ability . The system then picked which skills or strategies to use in the consequence , taking into write up its opposite ’s strength and weaknesses , just like a human table - tennis musician might .
Then , they pit their AI - control robot against 29 human race . DeepMind ’s robot subdivision beat all of the beginners and about 55 % of the intermediate musician , but it got trounced by advanced players . In an international military rank system , it would be a solid recreational player .
— 12 biz - switch moments in the history of artificial intelligence ( AI )
— ' ChatGPT moment for biology ' : Ex - Meta scientists grow AI model that creates protein ' not find in nature '
— AI models cultivate on AI - return data could spiral into unintelligible gimcrack , scientists warn
DeepMind ’s golem weapon system did have some taxonomic impuissance , however . For good example , it struggle with high balls and , like many of us , found backhand shots more challenging than forehanded ones .
Most of the human players seemed to like play against the system . " Across all acquisition groups and pull ahead rates , player agreed that playing with the robot was ' fun ' and ' engaging , ' the researchers wrote in the field .
The new glide path could be useful for a wide kitchen range of applications that call for nimble reception in dynamic physical environments , the researchers said .
