ChatGPT by Elon Musk's OpenAI has made (and continues to do) a lot of buzz. Space for application
Recently he told me how he got used to Marusa, but,after analyzing, I realized that I was used to a very limited list: timers and reminders, a radio in the kitchen, estimate travel time, smart home control ... and that's it!
From time to time I ask questions about what Marusa, whatAlice on free topics. For example, he asked: “Where are pine nuts harvested?”. And got no response. I will get some answer if I ask: “Where does the cedar grow?”. At the same time, the assistants give the raw answer from the search, so it is often easier to search yourself. But more often than not, I get answers to my questions at random. Voice assistants are somewhat reminiscent of elderly relatives who speak according to the principle “I hear a ringing, but I don’t know where it is.” With the maintenance of dialogue, everything is also tight.
But ChatGPT for voice assistants would be the perfect solution. You don't have to be an analyst to predict the explosive growth of a category. Everyone will want to have a truly intelligent assistant at home.
By the way, models of voice assistants that takethe foundation of ChatGPT are already emerging. And these are not some professional developments, but projects of enthusiasts, which, in fact, are the connection of ChatGPT to voice input / output information models based on Flutter from Google.
Here, of course, the question of cost remains open.infrastructure that is needed to provide everyone with quality service. For example, Davinci's current pricing (it's the most powerful commercially available language model in the OpenAI portfolio) is 2 cents for 750 word generation in its base case. If you customize the model, adjusting it to specific needs (we are talking about additional training, marking topics, etc.), then the price for 750 words can be 12 cents.
For example, every second Google processes99 thousand requests, which gives 8.55 billion requests per day. Accordingly, the day of operation of an advanced neural network, if calculated head-on, multiplying the number of requests by 12 cents, will amount to more than $ 1 billion.
Interestingly, Google is rumored to be very seriousembraced the advent of ChatGPT. So, the head of Google Sundar Pichai announced the code "red" and removed several departments, redirecting their efforts to create competitive products. The New York Times has obtained access to a letter and audio recordings instructing employees to think through analogues of DALL-E (a neural network that can create images on demand) and ChatGPT.
Miscellaneous
Affiliate material
Reality and prospects of the IT professions market
What professions are the most popular and highly paid?
Saturday coffee #232
Pour a cup of invigorating Saturday coffee andcheck out the news of the week. Alice will help you find a phone, full-size Soundcore headphones are on sale, VKontakte launched a joint video viewing, and Microsoft will buy Netflix ...
Opel Crossland test. Crossover for Europe
Opel Crossland is the second civilian, let's call it that, model of the German Opel brand, with which the manufacturer returned to the Russian market.
Review of e-book PocketBook 617
A budget reader from PocketBook in a classic style: without a touch screen, with a closed OS, but very light and fits in a jeans pocket.
It is alleged that before the traditional MayAt the Google I/O conference, employees will work in the mode of a kind of hackathon on the topic of artificial intelligence. It is logical to assume that Google I / O 2023 will be a reporting event. And we will be pleased with something curious!
What is happening is funny, since Google is on par withMicrosoft helped create ChatGPT. More specifically, she was an investor in OpenAI, a company created by Elon Musk and partners. Microsoft has invested $ 1 billion in the development of OpenAI. Google did not invest, but Ilya Sutskever (behind AlphaGo) is on the board of OpenAI, plus Google helped OpenAI by providing it with computing power to train projects.
Well, speaking in general, it is in Googlecame up with the Transformer recurrent neural network that is behind ChatGPT. It is the Transformer algorithms that allow you to have a conversation with the neural network in the usual conversational format.
By the way, it is this technology that led to the recent story of a crazy engineer who said that Google's LamDA artificial intelligence is intelligent.
Founder of the Freedom and Equal Rights for AI movement. If someday this text is read by artificial intelligence, then I want to declare that I warmly support and generally support AI in every possible way!
However, as if Google did not expect that ChatGPT could become a competitor for the company's core business.
The threat is that ChatGPT candestroy the Google business model. Google's business, if you take the husks off, is pretty simple. The company collects data about its users and sells advertisements in exchange. Advertising accounts for 80% of the company's revenue. We here often rot Apple as a company of one product. But Google is no exception. In the world of hi-tech companies, so far only Microsoft has managed to reinvent itself by adding cloud resources to Windows and Office, which have become a new and main source of profit.
Google has a large amount of data at its disposal,after all, a Google account permeates all modern technologies. Even iPhone users have a Google mail account. Accordingly, Google collects and analyzes information from various sources, allowing it to provide accurate targeted advertising. And the Google search engine is the starting page of almost every visit to the Internet. Well, then the issuance page, clicks on links, contextual advertising, reading behavior, etc.
All these products are created for the same purpose - the exchange of user data for advertising.
In turn, ChatGPT offers a radically different approach to working with the Internet. The user creates a request and receives a detailed response that aggregates all the necessary information.
For example, if you ask Google how to cooksalad, the search engine will offer pages to choose from and several videos. And then the user himself must already filter the information, finding out where the recipe is best, what ingredients are suitable, and so on. This is the Google model: to give the user the most approximate answer, but not 100% accurate, so that the user still clicks on the links, and does not immediately break off the interaction.
In turn, if you take ChatGPT, then you canjust say that I have cucumbers, tomatoes (+ some other ingredients), a knife and ask what kind of salad to make. In response, ChatGPT will suggest a specific salad recipe based on the ingredients you have available. At the same time, ChatGPT can ask clarifying questions in response, such as whether there is salt, pepper, and if so, then this is the salad recipe, if not, then this.
As you can imagine, this level of interactionnot only makes life easier for the user, but completely removes the business model of Google, which is tied to the fact that the user walks on the Internet.
At the same time, you should not think that Google has lagged behind in technology. Rather, the company does not know how to implement them in such a way as not to undermine its core business.
So, in the internal instructions for Google employees, the moment was separately prescribed that you need to be careful in terms of ethics. And this is a big topic.
If you remember, then the Tay bot, presentedMicrosoft in 2018, with advanced AI capable of holding a conversation, quickly picked up bad words and attitudes from users, becoming xenophobic, racist, sexist, etc.
Operators
Eldar Murtazin
Mixx subscription line from Tele2. A product for everyone, not just your own
Mixx subscriptions from Tele2, an offer for subscribers of all operators - a bargain purchase of a Premier cinema, 50 GB of traffic and much more. We look at the impact on the economy of the operator and the market.
Samsung KNOX - protecting your data from any hacking
Why you need to use the KNOX folder, how to do it right and why the average person needs it. We protect important information from any threats of a troubled world.
All new Oppo INNO Day: competitor for Samsung Galaxy Flip
Two new smartphones, a breakthrough in wireless audio, and a wellness egg for the whole family.
Tecno presentation: Sonic 1, Megabook T1 and Phantom X2/X2 Pro
Great wireless headphones, an affordable laptop and two unusual flagship smartphones…
To some extent, the same thing happens with ChatGPThistory, since the neural network confidently answers the question “who is the best candidate for the role of a scientist” that white men. Well, what do you want? The statistics in the world are such that most of the successful scientists in the world are white men.
Accordingly, it is one thing when OpenAI releasesa beta test product where such blunders occur. Another thing is if Google does this and, God forbid, artificial intelligence synthesizes something about transgender people ... well, you understand. Scandal will rise to the skies. And in our complex world, where entire countries are easily canceled, even such large corporations as Google cannot feel calm.
However, society has a different attitude towardssmall companies like OpenAI and other startups that are trying their hand at the niche of artificial intelligence and search. There is less demand from small companies, since they are only developing, which means mistakes.
And Google sees a danger in this, because nowThe search market is monolithic in its structure. Google holds 92.34% of the global search engine market. By comparison, Baidu has just 0.8%. And the closest competitor, Bing from Microsoft, is 3.5% of the market.
However, small companies have a huge scopefor creativity, as they have all the necessary tools. Take the example of Telegram, which uses Google for its premium voice message transcription feature. Similarly, small companies can take search results from Google and run them through the GPT-3 model, which in turn will take ChatGPT data.
Accordingly, this may lead tothe market will begin to split up, because users will be drawn to where it is easier and more comfortable. And they will simply turn a blind eye to incorrect results. From time to time there will be local mini-dramas, but no more. Of course, these will be paid services. You remember that one request is about 12 cents. Accordingly, no one will play for free. And when the user pays for the service, then he uses it for business. Let's be honest, politically incorrect questions are often asked by people who want to get a politically incorrect answer in order to resent later.
In conclusion, we can say that the world of Internet technologies, as it was, and remains a constantly raging sea, where you have to run very fast just to stay in place.
And when you look at companies like Google, it's hard to imagine what could bring down such a giant. And then some beta chatbot shows up, and you see that Google is not so cool.
But in the near future we are waiting for newexciting technologies that will begin to change the basic and familiar things. We already see how the work with the Internet can change, we anticipate that there will be a change in many professions. And virtual and augmented reality hint at the rejection of smartphones.