ChatGPT-4 Powered Bing vs Google Bard First Look Road to Artificial General Intelligence
Chat GPT Reached a hundred million active users in 2 months and became the fastest growing consumer app in history. This caused Google to feel an existential threat for the first time in years. As can give you direct Answers to search queries without needing to browse a list of potential results. Potentially killing Google’s ad business which currently accounts for 60% of their revenue in response. Google declare they called red and invited the co-founders back to Google.
Sergey Brin started coding again Microsoft saw this as a once in lifetime opportunity to take its share back in the search market. so they invested in CHATGPT and integrated into bing its long that dead search engine. This is forcing Google to roll out LaMDa it’s on chatbot based on a large language model in order to compete with Microsoft and OpenAI partnership.
Welcome to the war for the future of search. Microsoft has announced the integration of OpenAI GPT-4 into bing branding it as the NEW BIG. The AI enhanced bing has a new chat option in the menu bar alongside search which allows users to toggle between the two modes up on selecting the chat option users are greeted with a chat interface that provides three suggestions and a disclaimer which closely resembles the interface of chatGPT. In the search mode We can see results from bingchat next to the traditional search results. I saw some mixed opinions about it. Some people love it Some people hate it. I find it little bit distracting, but functionally is is great to have all this information from bing chat. Next to traditional search screen The new bing will have access to current information from the web and it will also clearly State its sources on the answers. It gives enabling users to verify the information bing is capable of handling complex tasks that usually require manual effort such as generating meal or travel plans and writing code.
The AI ability to comprehend and respond to natural language. Queries also means that bing can handle creative tasks such as writing a rhyming poem or creating a short story. Being can engage in a conversational exchange with users providing a more human like research experience. Google on the other hand responded with Bard a new experimental conversational service powered by LaMDa a large language model. Sundar pichai mentioned that bard is currently in external testing phase with Trusted testers. The word experimental Chosen specifically to prepare the public for the potential of wrong information or biased responses that may come from Bard. Bard’s large language model is a lightweight version of LaMDA that require significantly less computing power. It’s understandable why they choose a lighter model to start with. It has been estimated that cost of operating chatGPT is roughly 1 to 3 million dollars per month with each query requiring approximately 3 cent at Google serving 10 billion queries per day would result in an annual cost of 110 billion dollar which would not be Financially feasible to sustain. When we look at the user interface of bard It looks similar to chatGPT. The result page does not have sources or citations at least for now to provide feedback there are thumbs up and thumbs down button. The refresh button allows you to rerun your query while the check it button takes you to a direct or relevant Source on the web related to your question. Here is a quick comparison test between ChatGPT and Bard’s responses for the same exact prompt.
Which one do you think did a better job ? both companies take a hybrid approach combining a traditional search experience with chat. It makes sense since search is still good at head queries, like people, weather, locations or movies. The all time highest traffic for Google search was During the FIFA World Cup final as people search for scores.
Large language models such as ChatGPT excel at handling tail queries while also offering incredible creative opportunities. It makes sense to have a hybrid approach combining traditional search with Chad based quick answer. Some people prefer direct answer a quick response to what they’re searching for and some people like the research experience not the destination but Journey while researching for something they want to discover new information and new people and that’s totally fine.
It’s too soon to determine whether traditional search and chat based search will eventually merge. Only time will tell everybody asking the same question. where are we going with all these developments AI arms race between Tech Giants will accelerate further we will see better and more capable large multi models Combining not just written texts but also images , audio and video like PALM model from Google.
In the coming month the scalability of large language model will be an important issue given the messive computational resources require. That can only be supplied by Tech giants like Google or Microsoft. Another challenge lies in the need to continually and intelligently update these models with the latest information available online which is constantly changing. finding cost effective ways to do this will be an important challenge to overcome. effects on website Depend on search traffic will be catastrophic.
You will ask for recipe you won’t need to visit the recipe website It’s clear this will happen based on past experiences such as when Google introduced featured Snippets to search. Some websites reported a 20% decrease in traffic as a result. The former director of AI at Tesla Mentioned Github Copilot in his tweet saying that already. 80%, of code he is writing on a daily basis had been written by AI guess who owns GitHub
Microsoft it’s hard to call a multibillion-dollar company and Underdog but Microsoft is giving me that same feeling. I got When watching Morocco reached to semi-finals of the World Cup last year and I love underdogs. generative capabilities of these models Will grow rapidly until to a moment when it becomes self inventing itself. we know that before 2010 training compute grew in line with Moore’s Law doubling roughly every 20 months since the Advent of deep learning in the early 2010 the scaling of training computer has accelerated doubling approximately every 6 months Based on these facts it’s safe to say that eventually search engines will transition into answer engines.
Answer engine will transition into hyper Advance personal assistants accessible from are mixed reality headsets and augmented reality glasses. People will start hiring these models as employees. They never sleep never get sick or tired and they will be able to do knowledge processing work, better than us sooner or later in the future Specialized multi models that focus on individual domains or what I like to refer as artificial intelligence will begin to interact with each other as well as with data from robotics Solutions operating in both real world and simulated environments. This will be the point when discussions about artificial general intelligence will become much more serious and prominent Once we reach the AGI artificial super intelligence level won’t take too long to reach. Over the next decade I believe that cost of intelligence will rapid decrease to almost zero leading to democratization of intelligent and creative power. As a knowledge worker I reflect on the future of my career and acknowledge that it may not be relevant in 20 years. Despite this I remain optimistic and eager for the new opportunities that will come with the rapid advancement of artificial intelligence. I’m fully committed to embracing this change and exploring all that it will offer