We Don’t Need No Education? - Teaching Rules to Large Language Models through Hybrid Speech Governance

Artificial Intelligence doesn't know what's 'true'. Especially, generative AI models like chatbots veer from the truth, i.e. “hallucinate”, quite regularly. Chatbots simply invent information at least 3 percent of the time and sometimes as high as 27 percent. Given the (future) u...

Full description

Bibliographic Details
Main Authors: Wolfgang Schulz, Christian Ollig
Format: Article
Language:deu
Published: Max Steinbeis Verfassungsblog GmbH 2023-11-01
Series:Verfassungsblog
Subjects:
Online Access:https://verfassungsblog.de/we-dont-need-no-education/
_version_ 1811153856181043200
author Wolfgang Schulz
Christian Ollig
author_facet Wolfgang Schulz
Christian Ollig
author_sort Wolfgang Schulz
collection DOAJ
description Artificial Intelligence doesn't know what's 'true'. Especially, generative AI models like chatbots veer from the truth, i.e. “hallucinate”, quite regularly. Chatbots simply invent information at least 3 percent of the time and sometimes as high as 27 percent. Given the (future) use of such systems in nearly all domains, we might want such systems to follow more stringent rules of accuracy. And those truth-related rules are not the only rules for AI systems that warrant societal scrutiny. How those systems are trained will be crucial. In this blog post, we argue that a new perspective is key to tackle this challenge: “Hybrid Speech Governance”.
first_indexed 2024-03-11T11:38:58Z
format Article
id doaj.art-c4ba8bc15b7f44798ed258e6a9a69821
institution Directory Open Access Journal
issn 2366-7044
language deu
last_indexed 2024-03-11T11:38:58Z
publishDate 2023-11-01
publisher Max Steinbeis Verfassungsblog GmbH
record_format Article
series Verfassungsblog
spelling doaj.art-c4ba8bc15b7f44798ed258e6a9a698212023-11-10T11:44:49ZdeuMax Steinbeis Verfassungsblog GmbHVerfassungsblog2366-70442023-11-012366-704410.59704/2a94578e42e943d2We Don’t Need No Education? - Teaching Rules to Large Language Models through Hybrid Speech GovernanceWolfgang SchulzChristian OlligArtificial Intelligence doesn't know what's 'true'. Especially, generative AI models like chatbots veer from the truth, i.e. “hallucinate”, quite regularly. Chatbots simply invent information at least 3 percent of the time and sometimes as high as 27 percent. Given the (future) use of such systems in nearly all domains, we might want such systems to follow more stringent rules of accuracy. And those truth-related rules are not the only rules for AI systems that warrant societal scrutiny. How those systems are trained will be crucial. In this blog post, we argue that a new perspective is key to tackle this challenge: “Hybrid Speech Governance”.https://verfassungsblog.de/we-dont-need-no-education/AIADSAgenerative AIlarge generative ai modelsllmspeech governance
spellingShingle Wolfgang Schulz
Christian Ollig
We Don’t Need No Education? - Teaching Rules to Large Language Models through Hybrid Speech Governance
Verfassungsblog
AIA
DSA
generative AI
large generative ai models
llm
speech governance
title We Don’t Need No Education? - Teaching Rules to Large Language Models through Hybrid Speech Governance
title_full We Don’t Need No Education? - Teaching Rules to Large Language Models through Hybrid Speech Governance
title_fullStr We Don’t Need No Education? - Teaching Rules to Large Language Models through Hybrid Speech Governance
title_full_unstemmed We Don’t Need No Education? - Teaching Rules to Large Language Models through Hybrid Speech Governance
title_short We Don’t Need No Education? - Teaching Rules to Large Language Models through Hybrid Speech Governance
title_sort we don t need no education teaching rules to large language models through hybrid speech governance
topic AIA
DSA
generative AI
large generative ai models
llm
speech governance
url https://verfassungsblog.de/we-dont-need-no-education/
work_keys_str_mv AT wolfgangschulz wedontneednoeducationteachingrulestolargelanguagemodelsthroughhybridspeechgovernance
AT christianollig wedontneednoeducationteachingrulestolargelanguagemodelsthroughhybridspeechgovernance