Ready ChatGPT. When will neural networks replace affiliate marketers?

In this article, we are trying to figure out what and how an affiliate marketing specialist should do with neural networks: use it to work, temporarily ignore or be afraid of competition? We are happy to share with you what we have learned.

Artificial intelligence is evolving at a tremendous pace. Until recently, neural networks made only funny and ridiculous pictures, but today they can invent fascinating stories, compose music and conduct dialogues. A logical question follows: can a neural network also run ads and drive traffic?

As soon as this thought arises for the first time, it is no longer possible to “stop thinking”. The fear of losing your job due to competition with soulless robots is haunting at every turn, and a new day seems to be approaching the uprising of machines. No panic! Let's think logically.

The most popular neural networks for work

Let's not waste time explaining “what are neural networks for dummies” - you have access to Wikipedia. We will immediately proceed to discuss the current level of development of this technology.

Here are some interesting neural networks:

• Midjourney (generates images)
• Deep L (translator)
• Dall-E (a joint project of OpenAI and Microsoft, a simplified version of Dall-E mini is known as Crayon).
• WomboArt (image generator)
• Rytr (generates texts based on key phrases)
• ChatGPT (works with code and text).

Voice assistants Siri, Google Assistant, Amazon Alexa, etc. - this is also the brainchild of artificial intelligence of varying degrees of development, which can recognize voice, and speech, answer questions and execute commands. They can be improved if desired, as Mate Marschalko did, replacing Siri with GPT-3 to control a HomeKit with dozens of lights, thermostats, underfloor heating, an air handling unit, cameras, and more.

Moreover, speech recognition is already a milestone for LipNet, the first neural network that can read lips with an accuracy of 93.4%. Sounds like neural networks are “smarter” than some real people, right? Artificial intelligence is not lazy, does not forget, and does not “slow down” like a natural one ... It looks like an ideal employee.

Where are neural networks already working?

Neural networks are actively used to improve search results in search engines and social networks. AI-based recommendation algorithms for TikTok, Instagram, YouTube, and Netflix work.

The possibilities of using neural networks in advertising are priceless:

• Deep analysis and precise audience segmentation.
• Automatic creation of personalized ads.
• The ability to change the content of the ad dynamically, in real time.
• Increased conversions, and reduced advertising budget.
• Self-optimization of the campaign without human intervention.
• High-precision targeting.
• Processing numerous scenarios with hundreds of parameters in a short time.

As you can see, artificial intelligence can recommend, analyze and produce content. That is, the same as many flesh-and-blood specialists. And even the argument about the lack of a sense of humor and character in “soulless machines” is no longer relevant. After all, Siri is offended, and DeepText understands Facebook jargon. The DeepText neural network is constantly learning and will soon be able to recognize slang, words, and expressions with ambiguous meanings.

In fact, this is quite enough for routine tasks. Those who already use chatbots of the support service in correspondence with clients. And from such work as an assistant - a stone's throw to more complex duties, including launching advertising. Faster, cheaper and in the long run more efficient than people.

At the end of 2022, a student at a liberal arts university wrote a thesis using the ChatGPT neural network and defended it. While working on the thesis, he asked the neural network to draw up a plan, study the text and adapt it. In total, it took 23 hours to prepare the diploma. Anti-plagiarism rated the originality of the work at 82%. Now tell me honestly, how successfully did you write and defend your thesis?

In addition, The Guardian published an article written by GPT-3, and even a collection of poetry was published in Chinese. Some sci-fi novels and comics mention that they use the GPT-3 API for writing text.

Some areas of activity are closer to “contact” with artificial intelligence, while others are almost not affected by it. Therefore, opinions regarding the danger are very different: from the complete denial of risks to the regular use of neural networks to facilitate work. Translators, designers, PPC specialists VS neural networks at this stage are carefully looking at either an assistant or a competitor.

Copywriter VS neural network

I don't want to lay off the duty on a robot, but I couldn't control my curiosity and, of course, tested the writing abilities of artificial intelligence. And all the results turned out to be very weak texts, overflowing with water and stamps a la “qualified specialists”, “professional service” and “competitive prices”. Maybe I'm just not deep into the technology. But at the level of available services, this cannot be a full-fledged replacement for a human author.

Now I will say a somewhat pathetic phrase, but copywriting is not the generation of words and sentences, but the search and construction of meanings. And a neural network can “write” something, but by itself is not capable of a creative approach, analysis of USP and experience. In short, AI can compile but not think.

The person who sets the task needs to think. And if you calculate the time it takes to write the task and subsequent edits, then it’s easier to write these two sentences yourself. Therefore, in copywriting, neural networks (today) can be an auxiliary tool, but they cannot replace cool professionals.

Let's check AI in action

To confirm or refute, let's take a simple task: write a text for an advertisement for the Nutra-offer DeepHair Growth Oil (drops for hair growth). Let's turn to Rytr, an artificial intelligence-based text manipulation tool.

AI allows you to set the language, text style, its purpose and the number of options. We briefly “explain” the neural network what it is about, and we get the following result from it:

We got a well-coherent advertising text. Standardized one - but let's be honest, most of the promotional posts are like that. The neural network coped well with the creation of a promo description of the product.

Nuances begin when it comes to human perception. The product is suitable for hair on the head and face, it's true. But women are likely to be put off by the prospect of facial hair growth, and the human publisher understands this, so he would not list these properties in one ad, but would divide them by audience. The neural network does not understand this, so it needs to “explain”, change the task and/or edit the text.

Thus, we come to the same conclusion: a neural network can automate routine tasks, but it will not completely replace you. You still need to control and interfere with the process of artificial intelligence.

And so in any activity. For example, for a programmer, ChatGPT checks the code and helps to find errors. Previously, this took a lot of time, but now this process is automated. Can a neural network take the place of a programmer in a company today? Not in this format.

Question from a reader of the public:

Will the neural network replace copywriters soon?

— No, the neural network is unlikely to replace copywriters in the near future. While neural networks can generate text, they lack the creativity and language understanding of a human copywriter.

Important editorial note: this answer was written by the neural network ¯\_(ツ)_/¯

Where neural networks have already won

There are several examples where artificial intelligence really took people's jobs.

Cosabella abolished the advertising department and replaced all its employees with the Albert neural network. It automates advertising on Facebook, Instagram, Google Ads, sends letters and messages to customers, analyzes data, and optimizes campaigns.

Fruit producer Dole also resorted to the “services” of the Albert neural network. The company managed to increase fruit sales by 87%. Dole used ad placement recommendations given by a neural network based on parameter analysis. Albert found that smartphone user engagement was the highest and shifted the bulk of the marketing budget to mobile.

Alibaba Group has implemented AI that can write 20,000 lines of text per second for product descriptions on Taobao, Tmall, Mei. The seller inserts a link to the product page, selects the desired storytelling style. The bot itself writes the characteristics of the product, gives out a relevant description in Chinese.

But these are isolated cases. In general, AI is only partially used. But very effective.

The benefits of neural networks in sales

Most marketing services use artificial intelligence. According to Gartner, the number of companies using AI has grown by 270% since 2015, and this growth is increasing exponentially. Based on neural network technologies, targeting in advertising accounts of social networks, campaign and bid optimization, chatbots, and creative creation services work.

The neural network can easily be entrusted with the following responsibilities:

• Demand forecasting (estimate when a product or service will be required by customers, to ensure timely delivery).
• Sales forecasting (determine when and what the customer will buy with a high degree of probability).
• Chatbots help to establish contact with customers, increase their loyalty, and save time for managers and operators to answer the most common questions.

Affiliate managers use the power of neural networks every time they launch campaigns on Facebook, Google, and other platforms. The neural networks that CPA networks and affiliate programs are equipped with identify low-quality traffic and help fight fraud.

Gambling, poker tournaments, online casinos use artificial intelligence to study user preferences, analyze their actions, generate personal offers, study satisfaction, etc.

Get to the point

Okay, this is all exciting, but how do you apply this information to your daily work? Let's try to “unload” the designer and entrust the neural network with the creation of an advertising image.

Let's start simple - the Crayon neural network offers the option of free work, and we will test it. Let's ask AI to create a picture for a banner, an offer is a tool for male power. We write a request: “a man is happy because he satisfied his wife in bed.” You can see the result yourself, we probably refrain from commenting:

Tried to give a second chance with the request “beautiful young long-haired blonde with a package of a natural nutritional supplement in her hand.” Look:

There is the result.

Rock musician Nick Cave has dissected a song produced by the viral chatbot software ChatGPT “written in the style of Nick Cave”, calling it “bullshit” and “a grotesque mockery of what it is to be human”, writes The Guardian.

Designer Maria shares her experience using the Midjourney neural network. For the query “3d fairy tale bear with a pot of honey in his hands”, she got quite a nice result:

But Gilloute published on Reddit female portraits created by the neural network, which could well become not only banner creatives, but also models of fashion catalogs.

The author writes that the creation of the image took only a couple of seconds. However, the query here is much more complicated:

Prompt: (extremely detailed 8k wallpaper), ((halfbody portrait:1.2)), (water abstract:1.2), (concept art:1.5), A young beautiful topmodel woman from China, (Chinese fancy clothes), (colored water particles explosion:1.3), professional composition, natural, sharp focus, wide angleNegative prompt: fat, large breast, naked, nude, video game, anime, b&w, ((bokeh)), 2D, cartoon, 3D, illustration, poorly drawn, textSteps: 40, Sampler: DDIM, CFG scale: 10, Size: 512x704, Model hash: 8e194b43, Model: protogenInfinity_protogenX86.

That is, it turns out, as in the old cartoon: “To lose a day in order to fly in 5 minutes.” In order for the neural network to draw a picture in 2 seconds, you need to work hard with the formulation of the problem. And this is a lot of work and special skills.

So, it may turn out that instead of “taking away” jobs from people, neural networks will create a new profession for people - something like a “formulator of correct queries”. The demand for such specialists already exists and will continue to grow.

You can already read about this profession on LinkedIn, it's a fait accompli. So far, one can only argue about whether it will be a prestigious “profession of the future” or just an auxiliary job. In this case, they will pay for it several times less than the 3D designer/modeler earns. Nevertheless, it is precisely such tasks that take up a lot of resources in the production of games, films, and websites.

In other words, highly qualified specialists will not be left without work. There will simply be a division on the basis of immersion in the profession. This can be compared to the production of dishes. In the modern world, it is mass-produced on mechanized machines. However, manual work also exists and is valued much more expensively. Something similar will probably happen to digital artists.

How to use neural networks effectively?

Okay, if you're not ready to hire a prompt engineer yet, but want to automate some of your work, there are two options left. These are 1) the free and almost useless AI services we covered above, and 2) the useful but more difficult ones to use.

The first category includes the “draftsmen” from the previous section, as well as a few sites that a marketer can use:

ThisPersonDoesNotExist is a generator of photorealistic faces of non-existent people. It can be used for any creative without the risk of claims, because this person never existed, and the “photo” was generated by a neural network.

Looka is a logo generator based on the description of the company, field of activity and your tastes (through the choice of styles, colors and exemplary symbols).

ChatGPT is a trending artificial intelligence chatbot launched at the end of 2022 and capable of working in a conversational mode, writing code and texts, translating, and giving accurate answers taking into account the context. Within 2 months after the launch, the audience of active ChatGPT users reached 100 million people

How to use ChatGPT to arbitrate traffic

ChatGPT can do so many things that it's not hard to "suspect" it of the marketing ability. One of his latest achievements is AI Radio, run by two chatbots, Bella and Adam. The same neural network can draw up a business plan, write a poem in verse, or solve a complex mathematical problem. Or, for example, write code. Let's check:

We asked the neural network to write the html of the medical blog landing page (for the white page).

By the same principle, you can “order” the neural network to write a cloaca.

But not everything is so smooth. More precisely, not always. Here's a counterexample for you:

ChatGPT was asked to write a program to ensure that only targeted users go to Facebook ads.

In response, we received a sea of “water” and no useful specifics. Why? Because "what question - such an answer." The neural network responds very evasively to non-specific requests, like a negligent student in an oral exam.

Thus, we returned to the task of formulating queries. Without correct, very specific and unambiguous formulations, you will not get any practical benefit from AI.

Instead of conclusions

The question is when neural networks will learn to make commercially viable products. By "products" we mean texts, program code, pictures, 3D models, business plans and roadmaps for their implementation.

AI learns very quickly. This means that in 3 years there will be AI products that can replace the illustrator, graphic designer, UI and, possibly, UX. A copywriter and a good programmer are unlikely to be replaced. You need to write the code individually for each task, think about how to do it better and more optimally with the available number of inputs. The text also cannot be compiled from existing sources, because analytical work must be done before writing. AI does not know how to think, it does not have fantasy and critical thinking. Will they invent it? Given the algorithms of the AI, it is unlikely.

Roughly speaking, AI works according to references. There are a lot of references for illustrations and design, it is possible to compile something decent from this even with the minimum technical specification. To do quality work, you need to set quality specifications. Many customers do not know how to do this. Maybe experts will switch to the creation of technical specifications and the general formation of meanings. And, of course, you also need to control AI, it is not perfect yet. 

Neural networks can help make a plan or throw ideas, but it’s hard to imagine that a neural network made colleagues love their work, and suddenly everyone began to like bad service. In solving advertising problems, it is very important to understand the specifics of the region, trends, market conditions, GEO, target audience requests, situations, etc. You can find out or invent all this, but the neural network cannot (just assume).

Traffic lights have long replaced traffic controllers, but this does not cause public panic. Even a parking sensor in a car that squeaks when approaching other objects is already intelligence. 

The best thing that neural networks can do today is to make it easier for people to do routine work. We already see that the existing services really facilitate the work of many specialists. But neural networks will begin to occupy our jobs no earlier than grannies on the bench will discuss them. 

In the meantime, we continue to find the best offers, sources, and ideas for you. Aff1.com managers are always in touch and ready to share their experience!