ChatGPT's New Features: Real-Time Info, Privacy Issues, and Bias Concerns
OpenAI has brought back the browsing feature
to ChatGPT after temporarily disabling it in July.
ChatGPT can now access the internet to provide
users with up-to-date information, according to an announcement from OpenAI,
its parent company.
Previously, the chatbot was limited to using
data up to September 2021, which meant it couldn't provide real-time
information.
On Wednesday, OpenAI, which is backed by
Microsoft, shared on X (formerly known as Twitter) that this new update allows
ChatGPT to break free from the September 2021 data cutoff and tap into current
information available on the web.
OpenAI stated that the "Browse with
Bing" option is now accessible to users of the paid versions of ChatGPT,
with plans to make it available to all users in the near future.
OpenAI had previously tested a feature that
allowed users to access the latest information through the Bing search engine
within its premium ChatGPT Plus offering. However, this feature was integrated
in May and then disabled two months later due to concerns that it could enable
users to bypass paywalls.
This new integration operates in a manner
similar to Bard and Bing AI, the chatbots developed and launched by Google and Microsoft respectively.
Alex Hanna, director of research at the Distributed AI Research Institute and former research scientist at Ethical AI at Google, told Al Jazeera, "What OpenAI is trying to do is get in on Google's market share."
Privacy concerns also arose with real-time
information access. There were worries that the bot could encounter harmful
material, misinformation, and copyrighted content online and present it to
users, as previously reported by the BBC.
Additionally, users must enable their chat
history to allow the new browser plugin to function, which involves consenting
to share their personal data with the model.
Hanna cautioned, "People are at risk of
having their legal data scraped. Research has shown that with certain types of
prompts, certain kinds of private information have been leaked through these
different systems. We don't know what is being leaked just because of how
private these companies have kept their data sources."
OpenAI mentioned that the latest feature would
grant websites control over how ChatGPT can engage with them.
Hanna pointed out that ChatGPT can sometimes
display hallucinations, misinformation, and inaccurate information. She also
highlighted how AI tools and search engines can perpetuate racial and gender
biases, as evidenced by research, including Safiya Noble's book
"Algorithms of Oppression," which reveals disparities in search
engine results for queries about Black women and white women, exposing racism
and sexism within the algorithms.
Earlier this week, OpenAI introduced voice and
image features to the chatbot, generating mixed reactions online. While some
users celebrated the updates, others expressed concerns about AI becoming too
human-like. Some users raised questions about how these updates might affect
smaller AI startups and software engineers, while others discouraged the use of
ChatGPT, citing recent lawsuits against OpenAI alleging copyright violations
and intellectual property rights infringement.
Post a Comment