There is a fierce battle over the prices of AI tokens between tech companies around the world, especially China (Alibaba, ByteDance) and America. Companies have reduced their token prices by up to 90% to attract more customers. Basically, a token is the unit of AI that determines how much work the AI has done for you. Just as we see electricity measured in units, every AI response is measured in tokens. The longer the question you ask or the longer the AI answers, the more tokens you spend.
it works like this
When you write something and send it to AI, it breaks your entire sentence into small pieces, which are called tokens. Think of it this way: Just like small children are taught to read by joining words. In English, a word is often a fragment, but Hindi has a slightly different structure. In Hindi, vowels and consonants together form a word, so to understand a Hindi word, AI has to break it into 2-3 pieces. This is why the meter (token count) runs a little faster when AI is used in Hindi.
AI companies race to token prices
AI companies charge money for their services on a token basis. That is, the longer question you ask AI or the longer answer you get from it, the more tokens you will spend and the more money you will spend. To reduce this expense, competition among tech companies has increased nowadays. Now many companies are fixing package rates for 10 lakh tokens, so that big projects can be completed cheaply. Especially Chinese AI companies have reduced the prices of tokens significantly, making the use of AI cheaper than before.
This is exactly like how telecom companies compete among themselves to provide internet data at low cost. So, now the real competition in the world of AI is over token prices. Experts believe that just as cheap internet data led to the digital revolution, in the same way the use of AI will increase rapidly in every sector like education, health, agriculture, business and media due to cheap tokens.
03 times more tokens are used in Hindi language than in English.
Modern AI models can remember up to 2 million tokens.
AI’s game of memory and tokens
Tokens also determine the AI’s memory. Each AI model has a limit on how many tokens it can remember at once; This is called the ‘context window’. When the conversation becomes too long, new tokens are added and old tokens are removed. For this reason AI can forget old things. For this reason, tokens are directly linked to AI’s ability to understand and remember. In the future, only that company will be at the forefront in this field which can process more tokens at lower cost and consume less power. Therefore, big tech companies are now building AI factories where tokens can be processed at much faster speeds and make AI even more powerful.
