Google has issued a blog post and a detailed paper sharing its perspective on the growth of artificial intelligence (AI) and justifying why it is taking a slower approach when it comes to rolling out some of its newer ‘AI-based’ innovations such as its controversial LaMDA chatbot and Waymo self-driving car. The blog post is credited to James Manyika ( SVP at Google, Jeff Dean (lead of Google’s AI division), Demis Hassabis (CEO and co-founder of DeepMind, which is owned by Alphabet), Marian Croak (VP of Engineering at Google) and Sundar Pichai, Google and Alphabet’s CEO. Google’s ‘explanation’ arrives as AI becomes the big buzzword of 2023 thanks to OpenAI’s ChatGPT chatbot. Many have called it the end of Google’s dominance on search and reports have also talked about alarm bells going off at the tech giant due to ChatGPT. In the blog post, Google’s executives noted that it “is an exciting time in the development of AI,” and talked about the company’s approach to implementing the technology. The blog post goes on to address the issues of complexities and risks with AI and states that Google’s development of the technology will need to address these. “We also believe that getting AI right — which to us involves innovating and delivering widely accessible benefits to people and society, while mitigating its risks — must be a collective effort involving us and others….It is critical that we collectively earn public trust if AI is to deliver on its potential for people and society,” it adds. Meanwhile, the paper notes that Google is applying a “scientific method to AI R&D with research rigour, peer review, readiness reviews, and responsible approaches to providing access and to the externalization and use of our innovations.” It can be seen as an explanation for why Google is not rolling out some of its bigger AI products to the public. The paper mentions it is “performing continuous adversarial and related forms of testing and has taken a differentiated and careful approach to access and deployment of novel systems such as LaMDA, PaLM, and Waymo.” Google’s LaMDA chatbot is built on a similar large language model (LLM) and is capable of having intelligent conversations in the style of ChatGPT. LaMDA was in the news last year due to an engineer claiming the chatbot was sentient. Google had also showcased how LaMDA was being used to help writers with fiction writing, though it admitted that the chatbot was more of a helper right now. PaLM is another language model, which stands for Pathways Language Model (PaLM), which will be bigger than LaMDA, GPT-3 (ChatGPT is based on GPT 3.5) and will support translations to other languages. Still, PaLM has not yet rolled out. The same goes for LaMDA, which is not open to the public to users. One can only sign up for beta testing via its AI Kitchen app. Waymo is Google’s self-driving project, which is still being tested in the US. The paper also focuses in detail on complexities and risks of deploying AI tools and lists them out as well. Google also stressed that AI is already being used in its core products such as Search, Maps, Photos, Workspace and Android phones. It also reiterated that many of the core underlying AI technology powering other products was developed by them, listing out examples of “Transformers, Word2Vec, Sequence to Sequence Learning, Federated Learning, Model Distillation, Diffusion Models, Deep Reinforcement Learning, Neural Nets with Tree Search, Self-learning Systems, Neural Architecture Search, Autoregressive Models, Networks with External Memory, Large Scale Distributed Deep Networks, Tensor Processing Units.”Transformer, for instance, is the technology powering ChatGPT- GPT stands for Generative Pre-trained Transformer. The blog post ends by stating that Google’s executives are “excited about what lies ahead in 2023 and beyond as we get ready to share some new innovative experiences!” Does this mean Google will showcase new AI products later this year, perhaps at its annual I/O conference? We will have to wait and see about that.
