In an exclusive interview with Search Engine Journal, Yury Pinsky, director of product management at Google Bard, offers insight into the current state and future trajectory of Google’s experimental AI chat.
The interview offers a first-hand perspective on where Bard is currently, where Google hopes to take it in the future, and how Google is addressing challenges like potential bias and misinformation with this new AI tool.
User comments and usage patterns
Pinsky reported that initial feedback on Bard since its release has been positive.
He noted that people have quickly integrated Bard into their workflows to figure out how to best use it.
“We’re hearing that all over the world, people are eager to engage and collaborate with Bard. I think it’s pretty cool that we’re taking this journey alongside our users as we learn and figure out how to get the most out of generative AI together.”
Most people use Bard to develop concepts, program, and understand complicated topics, Pinsky says:
“In terms of common themes or more specifically how people use Bard, we’re seeing most people use it for writing, meaning finding the right words to use based on a user’s idea startup, work with an idea, code and help. people understand complex issues.”
Bard vs. google search
Regarding the relationship between Bard and Google Search, Pinsky emphasized that the two are different but complementary products.
“Bard and Search are separate products; in fact, we see Bard as an experience that is complementary to search. Bard can help increase your productivity, accelerate your ideas, and fuel your curiosity.”
Google Search has long been a tool for finding information, while Bard was created as a chatbot aimed at increasing user productivity and creativity.
Bard’s current focus is how it can encourage creative thinking, not just a basic search that users can do on Google.
Pinsky continues:
“While the generative capacities in the search [can] helping people on their information journeys, staying true to our strong foundations of information quality and connecting people to a variety of sources and perspectives. [is crucial].
For now, we’re excited and focused on how people use Bard for creative exploration, in a different way than how they typically look for information with Google Search.”
Bard’s strengths and areas for improvement
When asked about Bard’s strengths and weaknesses, Pinsky believes Bard is adept at being a creative partner, especially for writing professional correspondence and the like.
“One of the great advantages of Bard is that he is a very effective creative collaborator. So for tasks like creating drafts of [something like] a professional letter… Bard can help you find the right words.
Bard also helps users expand their ideas to come up with innovative solutions, Pinsky explains:
“Another way we see people using Bard is to generate ideas. Bard can help you from a starting point like thinking about a family trip to generating suggested places to visit, and with Extensions, generating options for flights and hotels.
I really think what Bard excels at is helping people build their ideas to creative conclusions.”
Like other AI language models, Pinksy acknowledges that Bard sometimes makes up incorrect information, known as “mind-boggling”.
It encourages people to use the feedback tool to identify inaccuracies, and highlights Google’s dedication to transparency and accountability in the responsible development of AI.
“In terms of improvement, we have been transparent that hallucinations remain a known challenge for LLMs.
When we launched Bard, we published a overview by James Manyika, our Head of Technology and Society, which examines many of these limitations and our approach to mitigating them.
We believe this transparency is important and fundamental to being responsible with generative AI.
So we encourage people to use the thumbs down button and provide feedback if they see a hallucination or something that isn’t accurate. This is one way Bard will learn and improve.”
Quality assurance and expansion
Pinsky discussed how Google constantly evaluates and improves the quality of Bard’s response through user feedback.
“Our user research teams spend a lot of time with our users to better understand the features they resonate with, how our responses can be better, and how they use Bard.
Plus, our ‘thumbs up and down’ feature lets us gauge how good Bard is with the answers and whether we’re making progress.”
He stressed that expanding access to Bard aligns with Google’s mission, but maintaining high-quality answers and responsible AI development remains the top priority over speed of deployment.
“When we announced Bard in February, we initially opened it up to our trusted testers before making it more widely available to the public. These testers help provide critical feedback before a wider release.”
Ethical measures for Bard
Addressing concerns about ethical issues such as bias and misinformation, Pinsky said Bard was developed according to Google. principles for ethical AI.
“We do our best to build all of our products according to our AI principles, where we have noted that we seek to avoid unfair impacts, including those related to… political/religious beliefs, race, ethnicity, gender, nationality , income. , sexual orientation and ability. We took the same approach with Bard.”
The team has taken steps to avoid unfair results related to sensitive attributes when using Bard. The system is regularly reviewed by people who provide feedback and evaluation. If any problems do appear, Pinsky says, Google can quickly take steps to fix them.
He continues:
“We use human feedback and evaluation to improve our systems, but like all LLM-based interfaces, Bard makes mistakes, and when we find that the experience doesn’t work in a way that aligns with our approach , we will work quickly to fix it.
More generally, finding ways to represent different points of view or prevent bias is something that society itself struggles with – it’s a very complex issue and one we continue to work on.”
Standing out in the AI market
When discussing how Google aims to differentiate Bard in an increasingly competitive space, Pinsky highlighted Google’s people-centric strategy to make Bard stand out amid increasing competition from chatbots powered by large language models (LLM).
“I feel very fortunate because it’s an exciting time to be working in this space – there’s a vibrant ecosystem with lots of options, which is great for consumers and the advancement of technology.
At Google, we like to say: put people first, and the rest will follow. And that’s exactly how we’ve approached and will continue to approach Bard, focusing on what people are telling us they want to do with the technology, and we think that’s going to be a key differentiator.
For example, people told us they wanted to debug code, so we added coding capabilities. People wanted a more visual experience, so now Google Lens can analyze your photos. And people wanted help solving math problems, so we made the model smarter in logic and reasoning.”
He says Google aims to tailor Bard to users’ needs by integrating it with popular Google apps and services. Pinsky believes this integration and focus on user experience will set Bard apart from other LLMs.
“Bard’s ability to integrate the Google apps and tools you use every day with Extensions is truly a game changer. And Bard is the first LLM to admit that it may not have all the right answers because, through the Google It button, people can check the results.
Overall, I think a key differentiator is that we hold ourselves to a high level of accountability in AI and are taking a deliberate and thoughtful approach when bringing new forms of AI technologies into the world.”
Bard and Google advertising business
When asked how Bard could affect Google’s ad revenue, Pinsky explained that Bard isn’t focused on making money from ads. Instead, the priority is to provide users with a positive experience interacting with Bard.
“Our focus right now is not on ad monetization, but on creating a great experience for people. And I want to reiterate our privacy commitment that people’s Bard conversations are not used to show you ads”.
The future of Bard
Pinsky discussed his vision of the future potential of generative AI as Bard. He believes this technology represents an exciting new frontier for innovation.
“We really think generative AI is the new frontier of innovation, especially as we bring its magical capabilities together with our products in a way that can really help people.”
While he didn’t provide specifics, it’s clear that Google is focused on improving Bard’s abilities and improving the system over time.
As Pinsky said:
“We’re still in the early stages of unlocking the potential of this technology, and I, for one, can’t wait to share even more features as we continue to develop in a bold and responsible way.”
Summary of the main conclusions
Here’s a summary of the key information we learned Pinsky during the interview:
People find Bard useful for brainstorming, coding, and learning activities. Bard is positioned as a creative enhancement tool, while Google Search remains the primary source for finding information. Bard’s strengths include his collaborative skills in drafting and exploring ideas. Areas for improvement include reducing factual inaccuracies and “hallucinations”. Google is taking a responsible approach with Bard, emphasizing ethical AI principles. Bard’s differentiators include optimization of Google products/services and unique features like the “Google it” button. Monetization is not the current goal – delivering user value is the priority. More capabilities and features are actively being developed.
looking ahead
Bard shows promise as a collaborative tool for creativity and productivity. However, quality improvements are needed. Maintaining a focus on responsible development, user value, and ethical AI principles is critical to Google.
If continued deployment goes well, Bard could become a useful AI assistant. But it remains an experimental technology, meaning cautious optimism is warranted as capabilities evolve.
Featured image: Ascannio/Shutterstock
[ad_2]
Source link
