- i24news
- Middle East
- Technology & Science
- Check out Kosher.Chat, ChatGPT's Jewish-themed competitor
Check out Kosher.Chat, ChatGPT's Jewish-themed competitor
An ultra-Orthodox rabbi created a chatbot that would filter content that contradicts Jewish teachings
Rabbi Moishy Goldstein, a Hasidic Jew from New York, has just launched Kosher.Chat, an AI chatbot that works like ChatGPT, but with answers tailored to Orthodox Jews and based on Halacha (Jewish law).
Coming days after a decision by the rabbinical court of the Skver Hasidic movement to ban the use of artificial intelligence, Moishy Goldstein told the Jerusalem Post about his motivation to create a Jewish alternative.
According to the rabbis, "AI chatbots are tantamount to unfiltered internet and may answer questions ‘as a matter of fact’ in direct opposition to Torah's views and values," Goldstein said.
“The purpose of the bot is not to answer Torah questions... rather it is meant to be viewed as a mundane chatbot, used to answer mundane questions, but with a Jewish filter to prevent the answers from opposing Torah values.”
Goldstein added that while the bot is not meant to offer guidance on Jewish law and should never be used as a substitute for rabbinical council, a user can ask questions about the Torah "and can even get some pretty good answers," so long as those responses are treated with a grain of salt.
Using Kosher.Chat is similar to using other types of AI-powered chats, though the returned answers reveal the Orthodox values underpinning the bot. Asked about the meaning of LGBTQ+, Kosher.Chat said that "according to Torah values and Jewish law, homosexuality is considered a violation of the prohibition against same-sex relationships," while adding that it is "important to treat all people with kindness and respect, regardless of their sexual orientation or gender identity."
Explaining what sources the chatbot uses to provide its answers, Goldstein explained, “the knowledge base includes whatever the base OpenAI model knows from its own training, plus any additional information that I train it with, based on users' feedback."