|
Listen to this story
|
OpenAI’s conversational chatbot ChatGPT, which took the internet by storm, is probably the most significant development in the world of AI this year.
Within five years, nearly a million different users were conversing with ChatGPT, which can—besides answering general queries—explain codes and scientific concepts, write basic academic essays and even scripts for romcoms.
But ChatGPT does have its own limitations, some of which are quite well known. Last week, Writesonic introduced Chatsonic, with the tagline: ‘Like ChatGPT but with Superpowers’.
According to the creators, Chatsonic can do everything ChapGPT does and even more. So, is Chatsonic better than ChatGPT? Let’s find out.
Chatsonic vs ChatGPT
ChatGPT is built on OpenAI’s GPT-3.5 architecture of large language models (LLMs) and is fine-tuned with both supervised and reinforcement learning techniques.
Even though ChatGPT is impressive, it does not have the ability to connect to the internet. Its knowledge is limited till Q3 2021 and can only answer from what it has learned from the pretrained corpus.
If you ask ChatGPT about Argentina’s triumph in the recent FIFA World Cup 2022, it won’t be able to respond to it. In Sam Altman’s own words, founder at OpenAI, ChatGPT is incredibly limited.
In contrast, Chatsonic’s knowledge is not limited; in fact, it can surf the internet and converse with you on current events like the FIFA World Cup, for instance.
Chatsonic is not bound by the same limitations that plague ChatGPT. “ChatGPT is limited by the inputs or training data that cannot be updated so frequently. Whereas, Chatsonic is an advanced and powerful version that is able to keep up with current events, giving it an advantage in terms of accuracy and reliability,” according to Writesonic, the website behind Chatsonic.
When we asked Chatsonic about Apple’s self-driving car, it said that Apple has recently scaled back plans for a fully self-driving vehicle, which is indeed true.
In addition, when we asked Chatsonic if it’s better than ChatGPT, it said, “I’m not sure if I’m better than OpenAI’s ChatGPT, but I’m certainly an advanced chatbot that is capable of answering questions and providing information”.
Further, ChatSonic is built on the GPT-3.5 architecture. “It is also enhanced with additional proprietary and open-source models and algorithms. We do lots of preprocessing and postprocessing on top in order to be able to feed real-time information from Google, come up with contextual responses, generate images and understand user intent,” Samanyou Garg, Founder and CEO at Writesonic, told AIM.
Chatsonic also generates art
Another impressive feature that Chatsonic brings along is its ability to generate digital artwork. After ChatGPT, the next big development in the world of AI, would be the advent of generative AI.
AI models such as DALL-E2, Stable Diffusion and Midjourney are driving the generative AI revolution, which is only expected to improve.
The open-sourced text-to-image AI Stable Diffusion, by Stability AI—which also took the internet by storm—is now integrated with Chatsonic.
Now, users can chat with Chatsonic and ask it to simultaneously generate an image. Features of ChatGPT and Stable Diffusion at one single platform is indeed impressive.
How accurate is Chatsonic?
One limitation that LLM-based Chatbots have is that they are not always accurate. Meta’s Galactica AI was taken down in just three days for generating hallucinating and highly inaccurate results.
ChatGPT, in certain cases, also produces inaccurate results. Recently, Stack Overflow, the popular programming forum, banned all answers created by ChatGPT, citing a high degree of inaccuracy in the bot’s responses.
So, since Chatsonic is connected to the internet, does it solve the problem of inaccuracy? Well, that is what Writesonic is claiming at least. “Looking to generate factual content, including real-time topics, quickly and efficiently? Chatsonic can be your go-to AI tool,” it says.
Giving it the very ability to surf the internet means it can look for the right answers on the web for any given prompt. However, one reason ChatGPT is not connected to the internet is that it could lead to biases.
In fact, in a blog post, OpenAI claimed that ChatGPT is not connected to the internet and it can occasionally produce incorrect answers. Exposure to harmful content on the web could influence the chatbot’s behaviour. Since Chatsonic is exposed to the internet, there is a possibility that biases could creep into some of the results it generates.
“In order to tackle this, we connect with Google’s Knowledge Graph to supplement the GPT-3.5 model with real-time factual information on a given topic. This helps us ensure that the responses are accurate and up-to-date and also reduces the risk of the chatbot being exposed to harmful content. Of course, AI is still in its infancy, and we are constantly working on improving the accuracy and robustness of the system.
“Additionally, we take several steps to ensure our chatbot is safe from any malicious or inappropriate content. We use advanced language processing algorithms to filter out any offensive words or phrases. We also implement a robust content moderation system that screens all incoming messages for any inappropriate content,” Garg said.
The real Google killer?
When OpenAI made ChatGPT open for all, one of the most popular consensus among users was that it could replace ‘Google Search’.
Many believe that if ChatGPT is allowed to connect to the internet, it could be what ultimately kills Google Search. Some even labelled ChatGPT as the ‘Google Killer.’
But, could Chatsonic emerge as the real Google Killer? While it would still be premature to say that a bot like Chatsonic could possibly kill Google Search, nonetheless, it has successfully rattled Google.
Several at Google fear being left behind in terms of AI development. According to a recent report by the New York Times, Google has pressed the ‘code red’ panic button and reassigned teams to respond to the threat posed by ChatGPT.
However, Garg is of the opinion that Google Search will not be killed by chatbots like ChatGPT and ChatSonic. “Google Search is still the go-to for more complex queries and tasks and is still the most powerful search engine because it indexes billions of web pages, which would be very costly to do with an LLM,” he added.
It is to be noted that Google announced LaMDA, its own conversational bot, in 2021. Google has the resources at its disposal to respond to the threat posed by ChatGPT. Connecting LaMDA to the internet could possibly be a better alternative to Chatsonic.
However, the question is will Google do it? So far, it has not made its AI models like LaMDA or text-to-image model, Imagen, accessible by all for various reasons. But with Google pressing the panic button hard, we could expect some serious development in this field from Google next year.
