There are more and more negative issues with Google Bard, and internal employees are questioning: Is the massive investment worth it?

In the past few months, Google's parent company Alphabet has invited heavy users to continue internal discussions on the development of its AI chatbot Bard; some Google product managers, designers and engineers have questioned the effectiveness and functionality of AI, and even questioned whether the company's investment in massive resources to develop AI is worthwhile.

A senior product manager at Google Bard said on the forum Discord in July this year that for those AI chatbots based on large language models (LLM), "My principle is that unless I can independently confirm the accuracy of the information, I will not trust the results generated by the large language model. When the model reaches this level, I will certainly be happy to use it, but it has not been reached yet."

A user experience team director of the Bard project said in August, "I am still thinking about it. The biggest challenge facing large language models is what can they be used for, or what help can they provide? When can it have a significant impact on humans? Obviously the time has not yet come."

In response to ChatGPT launched by Microsoft's OpenAI, Google launched Bard in March this year. Since then, it has continued to add new features to Bard, including AI analysis of pictures, multi-language services, etc. Last month, Google took a big step by integrating Bard into its core products such as Gmail, Maps, Docs and YouTube.

However, Google has also begun to increasingly face the troubles caused by AI, including fabricating facts and providing users with potentially dangerous suggestions. On the day Google announced on the 19th of last month that it would integrate Bard into its core products, Google also launched a one-click search function in Bard to help users check whether the responses generated by AI are true.

For Google, ensuring Bard's success has become the company's top priority. Although Google is far ahead in the search field, with 80% of its parent company's revenue coming from this, with the explosion of generative AI models, Google's dominance in the search field is being challenged, especially OpenAI and a number of start-up AI companies, which may end Google's dominance.

Internal forum messages show that even the R&D director of the Bard project is confused and conflicted about the potential of this product. A product manager suggested that Google should restrict users from using Bard in "creative, brainstorming programs," but that using AI programming is a good choice. After all, coders ultimately need to verify whether the code works.

Google said in a statement that discussions about Bard's limitations and potential in internal forums are routine and not surprising because they are part of product development. "Launching Bard is an experiment, and we are eager to hear feedback from users so that we can improve the user experience."

Moreover, discussions on the Discord forum not only involve the function and effectiveness of AI, but also involve international politics and labor ethics. Google and Amazon have participated in contracts to provide AI tools to the Israeli military. In mid-July, a forum user questioned Google's use of AI as a lethal weapon. The user was immediately banned from the forum. The administrator stated that the forum prohibited discussion of politics, religion, or other sensitive topics. In the same month, another user questioned Google's use of low-paid, overworked contract workers to optimize Bard. Google said that manual optimization is an important part of Bard's research and development, otherwise the AI ​​may make major mistakes.

Some forum users questioned the huge amount of resources invested in developing Bard. One user wrote, "Are there any efforts to reduce the massive resource investment in large language models? Especially the huge amount of water and the need to use GPUs on a large scale." In this regard, Bard user experience director said that developing Bard is like chip design or supercomputer, "I believe we will find a way to reduce resource investment and achieve the same effect."

In fact, not only is Bard internally questioning the value of AI, its competitor ChatGPT is currently experiencing similar problems. Wall Street News previously reported that the latest data shows that its growth rate has begun to show signs of stagnation. More and more users have not brought corresponding economies of scale to ChatGPT, but have continued to increase costs. Silicon Valley giants can only try to charge users higher value-added fees, causing user dissatisfaction.

Sequoia Capital said that the biggest problem currently facing generative AI is not finding use cases or needs, but the inability to prove its value to users. Some believe that as the enthusiasm of investors and users fades, a group of obsolete players may emerge in the field of generative AI next year.