The world of fairy tales should be a matter of movies and books, but when it comes to artificial intelligence, dubious “AI” companies and ill-informed bloggers become veritable fabulists. The truth is that AI will change the working world, but many statements about artificial intelligence definitely belong in the realm of myths and fairy tales.
Contrary to popular belief, an artificial intelligence does not understand what it is doing. If you, for instance, ask your Alexa speaker to play relaxing music, it doesn’t know what music or relaxation is. Alexa simply scans the keywords in the music streaming services it is linked to. Then it plays the playlist that best matches the “relaxing music” search query. Alexa does not know whether the music you are listening to is really relaxing music.
Although they are very complex, procedures such as a text translation from English to German do not require an understanding on the part of the AI for both languages. On the contrary, the software determines language patterns in the text and uses them to produce the statistically most suitable translation.
(More information on machine translation of texts can be found on Linguatec in the article “Machine Translation“.)
The basis of artificial intelligence is an algorithm that controls its “behavior”. Artificial intelligence will never make changes to it – that would be like pulling oneself up by one’s own bootstraps.
Note: The reason, for example, why Google Translator translations are gradually improving is because the company adapts the algorithms or develops completely new ones. (Read more about this on Google AI Blog)
Also, the AI does not get any smarter than its training data when in use. In fact, the opposite is true. For instance, an AI sorts a company’s inbox by categories: Invoices, reminders and purchase orders. It was trained with data for these categories. However, now it has to deal with a complaint. The AI finds the most similarities in the order category. Therefore, the AI will classify all complaints as orders in the future. If other elements such as revocation or inquiries are added, the AI will make even more mistakes. Artificial intelligence must, therefore, be retrained whenever the data to be processed changes.
This also explains why an AI is always only as smart as its training data. If you use badly prepared data, the software will also work poorly.
Tip for the generation of training data for AI:
At clickworker you can order the new creation, editing or evaluation of thousands of training data for AI according to your needs.
Machine learning is a subfield of AI. Other subareas are robotics or automation. However, machine learning and AI are currently used synonymously, because impressive results are achieved with ML.
This assumption is also incorrect. On the contrary, the quality of the training data determines how well the AI will subsequently work. This can be achieved by clearly specifying the goal, i.e. what the AI is expected to do. Simply feeding an AI with large amounts of data and then assuming that it will solve all the tasks it is given, also belongs in the realm of myths.
The biggest myth surrounding AI is that it is objective. As previously mentioned, the decisions of the software depend on their training data – and this training data is developed by human beings. In 2017, the “hotness filter” of the “Faceapp” app made the headlines. This filter was supposed to automatically make selfie photos “hotter”. The problem was that it had only been trained for “hotness” with Caucasian model faces. Therefore, the filter gave people with dark skin a much lighter complexion, it worked so to say racist.
The example mentioned above was only “annoying”, but the one-sidedness of training data can have very serious consequences. For instance, a badly trained algorithm will only grant loans with bad terms and conditions to certain groups of people, or it increases health insurance premiums or ascribes criminal characteristics due to prejudices in the training data.
(To learn more about the incredible biases in the training data of image recognition systems, read this article: Excavating AI)
According to a study by the US think tank Brookings Institution, AI will not, as is expected, replace “simple” professions. Instead, it will affect the jobs where an academic education is required. This is because the strength of artificial intelligence is to recognize patterns and evaluate data volumes extremely quickly. And many academic professions are busy doing just that. For instance, financial advisors, engineers or sales managers, just to name a few. Manual trade sectors, on the other hand, are hardly threatened by AI; and it is hardly likely that professions in the social sector can be replaced by artificial intelligence.
(More information on the threat AI presents to the professional world can be found in the study of Brookings “What jobs are affected by AI? Better-paid, better-educated workers face the most exposure“.)
As explained in item 1, AI does not understand what it is doing. Therefore, consciousness cannot be attributed to artificial intelligence. However, there is little doubt that AI researchers dream of creating an artificial intelligence that has a consciousness. But at the moment the “I” in artificial intelligence stands for algorithmics.
(Interested in the topic? Wired examines this in detail in the article “Will AI Achieve Consciousness? Wrong Question“).
Believing in an AI consciousness is also linked to the myth that the AI is out to destroy mankind. But since self-conscious systems are still a long way off, if they can be realized at all, these fears are undeniably reserved to the world of film and literature.
Let us know what you think. Is there a myth missing or are our assessments incorrect? Write to us or post a comment. We look forward to your feedback.
PS: This text was written by a human being.
No related posts.
Dieser Artikel wurde am 12.February 2020 von Thomas geschrieben.
We are using cookies to give you the best experience on our website.
Find further information in our data protection policy. Change cookie settings.
Cookies are small text files that are cached when you visit a website to make the user experience more efficient.
We are allowed to store cookies on your device if they are absolutely necessary for the operation of the site. For all other cookies we need your consent.
You can at any time change or withdraw your consent from the Cookie Declaration on our website. Find the link to your settings in our footer.
Necessary cookies help make a website usable by enabling basic functions like page navigation and access to secure areas of the website. The website cannot properly without these cookies.
If you disable this cookie, we will not be able to save your preferences. This means that every time you visit this website you will need to enable or disable cookies again.
Any cookies that may not be particularly necessary for the website to function and is used specifically to collect user personal data via analytics, ads, other embedded contents are termed as additional cookies.
Please enable Strictly Necessary Cookies first so that we can save your preferences!