This calls for a high degree of flexibility in language use. It also shows that their algorithms are indeed more advanced in this area. Obviously, this is not an algorithm that could have been trained over a long period of time. Currently, we don't have backend data, so we don't know the specific computational power consumption.
But one thing, I feel that everything we're doing may become meaningless."
A female engineer seated to the left of Philip Hacks was the first to speak up.
Philip frowned and glanced at Nicole, who had spoken.
Though the words were unpleasant, the truth seemed to be just that. At least in terms of language models, they had lost.
However, this obviously wasn't what he wanted to hear.
Because this meeting was about solving problems, not just complaining.
In February this year, when Microsoft invested an additional ten billion US dollars into the company, the valuation had reached 29 billion. After half a year's development, the valuation had grown to ninety billion.
The cooperation with Apple was supposed to give the company's valuation a chance to surpass a hundred billion and secure its position as the world's third most valuable startup, but it was clear that Dou Dou's sudden emergence had turned that hope into a pipe dream.
This was evident from today's stock market trend.
Microsoft and Apple both took a near-identical trajectory; starting high and then trending downward.
Especially Apple, which suffered a true "black hour" in the afternoon, plunging from the daily high of 168.23 US dollars to the closing price of 122.36 US dollars.
There was no doubt this was an extremely terrifying figure,
The plummet of these two tech giants also directly affected the entire tech sector.
The meteoric rise of Youwei and the emergence of Dou Dou had made sensitive Wall Street pessimistic.
Things shouldn't have been .
The pressure almost moved at the speed of light onto these tech workers.
This was a terrifying thing.
If Huaxia's technology truly demonstrated an overwhelming advantage, it would trigger a series of reactions.
The greatest impact, of course, would be on the currently most successful large language model, ChatGPT.
New startups could choose to use the latest algorithms and start from scratch to design their own large language model architectures. Companies like Apple, and Google... But if they abandoned their current architecture, it would mean that all their efforts over the past decade had been in vain.
Everyone would be back at the same starting line.
The world of capital is ruthless. When ChatGPT loses its technical leading edge, and its earning capacity plummets, investors will abandon it without hesitation.
It is sad because aside from staying ahead, they have no other options.
Admitting defeat would mean a fleeting glory.
"Nicole, perhaps what you say is true, but you should know that our meeting is to solve problems. Let's buck up, everyone. We should recognize that maybe the other side has indeed gained some advantages, but we also have our strengths, right?" Philip said gravely.
Philip Hacks's words finally started to liven up the originally disheartened atmosphere in the meeting room.
"Yes, Nicole, I think Philip is right. Why did the Huaxia people only showcase this technology on a social media platform? I think there's a reason for that. I believe their technology is not capable of multi-modal processing, which should be our biggest advantage at the moment. This is something that needs more time to accumulate," one of them said.
"Right, in terms of analyzing complex images, understanding, and giving sensible advice, this is our greatest strength. Dou Dou hasn't shown this capability, the other side is playing to their strengths and avoiding their weaknesses. I've read Qiao's papers, the integration of causal relationships does indeed make their algorithm more advanced in understanding natural language.
But that's pretty much it. I believe that Dou Dou doesn't have the capacity for structured data extraction or directive model analysis. Plus, in such a short time, they couldn't have possibly conducted complex reasoning chain model tests, which is where our advantage lies," another voice chimed in.
"I agree with George's point of view, and I also think that the reason the Huaxia people launched such an unorthodox test is most likely because they lacked confidence in passing the security department tests. We should amplify this advantage," another person added.
Philip Hacks listened carefully to everyone's discussion.
Well, at least the team's cohesion hadn't completely dissipated.
Many suggestions were sensible.
In dealing with investors, they could delve deeper into specialized fields. Tasks such as precise identification of professional medical images and providing rational suggestions, strengthening the application of computer vision...
These functions were already being tested, and although they weren't yet at the stage where they could be launched as applications, they were areas where they held a technological lead.
And clearly, the other side had not yet started researching in this area.
Now, what they needed to do was to let the bigwigs on Wall Street know that the large language models were not just about complex natural language processing, but also involved a series of applications surrounding computer hearing, vision, and so on.
The emerging Huaxia technology would undoubtedly be lagging behind in this aspect.
He hoped so and, logically speaking, it should be the case.
"Alright, I've taken note of everyone's opinions. Next, I will propose to the board to focus on the development of multi-modal functions, which may very well become our next key direction. Everyone, get ready. I'd like each of you to submit a plan by the day after tomorrow at the latest for me to review. Understood?"
"Also, I suggest that all of you think carefully about how we can integrate more advanced algorithms into our product! Okay, that's all for today." Philip Hacks concluded the meeting.