Click the card below to follow the " CVer " official account
AI/CV heavy dry goods, delivered in the first time
Click to enter —> [Target Detection and Transformer] Exchange Group
Cressy sent from Aofei Temple
and reprinted from: Qubit (QbitAI)
Jacob Devlin , one of BERT's works, joined OpenAI only 3 months ago, and was revealed to have returned to Google.
Devlin returned to Google and worked closely with Bard lead developer Slav Petrov, according to people familiar with the matter.
The reason why Devlin left Google before was related to Bard, which can be described as "come to Bard, go to Bard".
The Information directly used "AI boomerang" to describe this incident.
Regarding this return, some netizens also speculated that this may be a strategy for Devlin to seek a salary increase.
From Google's point of view, Devlin has been harshly critical of its executives before.
The return of Devlin this time "regardless of previous suspicions" can also be seen that the competition for AI talents is too fierce.
Dissatisfied with Bard's "plagiarism" and choosing to leave
The trigger for Devlin to leave Google was related to the chat robot Bard.
A few months ago, it was revealed that Google Bard used the content in ShareGPT as training data.
The latter is a sharing platform where users can share content chatting with ChatGPT.
If the rumor is true, it is equivalent to Google training Bard with ChatGPT.
According to OpenAI's user agreement, the content generated by ChatGPT cannot be used to train other AIs, but the training results cannot be used for commercial purposes.
Obviously, Google's approach did not comply with this regulation.
After hearing the news, Devlin directly approached the CEO, Brother Pichai, and issued a warning.
After Devlin's warning, Google did stop the move.
In addition, it is also reported that the training data previously involved in the use of ChatGPT has also been quietly deleted by Google.
However, Google refused to admit that it had used the data in ShareGPT for training.
Bard's answer to the matter was more nuanced.
The answer to the direct inquiry was of course no, but Bard "admitted" that he had browsed the content in ShareGPT.
Devlin, a programmer, was already dissatisfied with using data from other companies to train his own products, and Google's response was even more difficult to make people—at least Devlin—satisfied.
As a result, before Bard was officially released, Devlin left Google and turned to OpenAI.
Progressive Bard
Since its birth less than half a year ago, the performance of Bard has indeed gradually improved, and its functions have become more abundant.
Viewed from the side, the reaction of netizens also proved this point:
In the internal testing phase and early days of Bard, even Google's own employees felt disgusted.
The newly released version even made some netizens regret paying for ChatGPT.
As for Bard itself, there have been major updates in recent months:
First, in April, Bard added a series of code-related functions, including generation, interpretation and debugging, and can also link colab with one click.
At the Google Developers Conference three weeks later, a number of heavy updates were launched.
First of all, there is no need to wait in line for the test qualification, and you can use it directly.
There are more languages supported, including Chinese, there are more than 40 kinds in total.
This version of Bard not only integrates the newly released PaLM 2, but also integrates other Google applications.
For example, with the help of image recognition tool Google Lens, Bard also has "visual ability".
In the content of the reply, you can also ask Bard to add pictures.
Not just image recognition, online documents, network disks, mailboxes...even maps are integrated into the new version of Bard.
In addition to its own products, Adobe Firely, which people love to see, has also been "invited" into Bard.
I don't know if Bard's improvement is the reason why Devlin chose to come back, what do you think?
参考链接：
[1]https://twitter.com/glenngabe/status/1672284772721209344
[2]https://www.theinformation.com/articles/ai-boomerang-googles-internal-critic-returns-from-rival-openai
[3]https://mp.weixin.qq.com/s/L2VKm3y3HLyB0GCqoM8NmA
[4]https://mp.weixin.qq.com/s/4FIe56KhaPdef9y_HswxFA
Click to enter —> [Target Detection and Transformer] Exchange Group
The latest CVPR 2023 papers and code download
Background reply: CVPR2023, you can download the collection of CVPR 2023 papers and code open source papers
Background reply: Transformer review, you can download the latest 3 Transformer review PDFs
目标检测和Transformer交流群成立
扫描下方二维码，或者添加微信：CVer333，即可添加CVer小助手微信，便可申请加入CVer-目标检测或者Transformer 微信交流群。另外其他垂直方向已涵盖：目标检测、图像分割、目标跟踪、人脸检测&识别、OCR、姿态估计、超分辨率、SLAM、医疗影像、Re-ID、GAN、NAS、深度估计、自动驾驶、强化学习、车道线检测、模型剪枝&压缩、去噪、去雾、去雨、风格迁移、遥感图像、行为识别、视频理解、图像融合、图像检索、论文投稿&交流、PyTorch、TensorFlow和Transformer等。
一定要备注：研究方向+地点+学校/公司+昵称（如目标检测或者ransformer+上海+上交+卡卡），根据格式备注，可更快被通过且邀请进群
▲扫码或加微信号: CVer333，进交流群
CVer计算机视觉（知识星球）来了！想要了解最新最快最好的CV/DL/AI论文速递、优质实战项目、AI行业前沿、从入门到精通学习教程等资料，欢迎扫描下方二维码，加入CVer计算机视觉，已汇集数千人！
▲扫码进星球
▲点击上方卡片，关注CVer公众号It's not easy to organize, please like and watch