“The eye is clean, youthful, attractive. If the woody vanilla has a free hand, the almond and hazelnut owe nothing to the barrel. Of a roundness without excess, it is a 2003 preferring – we are grateful to him for it – finesse to power. It has not sunburned and its general harmony is in the good average. It is believed to be subtle and balanced enough to reach three to four years.
Yes this kind of criticism thrills you, it’s probably because you love words as much as wine. But take care, explains Scientific American: it is no longer certain that this type of text was indeed written by a human being.
The site resumes results published in the International Journal of Research in Marketing by an American scientific team, which has developed an artificial intelligence capable of analyzing a wine (or a beer) and drawing a truer-than-life description.
The announced objective is to work in “complementarity with humans for online reviews”, i.e. helping wine and beer professionals effortlessly produce blurbs for their beverages, or giving humans templates to build on. The research team also says that other types of products, such as coffee or cars, could also be the subject of such an initiative.
This artificial intelligence developed according to the principle of machine learning – he was made to swallow 125,000 wine reviews and 143,000 beer reviews – seems capable of producing texts similar in all respects to those written by humans, which would then become strictly useless. Tests have indeed shown that it was almost impossible to distinguish the reviews written by this AI from those which were not.
Still, the tool only works based on existing reviews: at no time is there any question of having them taste the wine or beer they have to review. Its role is simply to compile the reviews found in its database and format them nicely. To this day, the human nose, eye and palate therefore remain irreplaceable. But Praveen Kopalle, marketer and co-author of the study, says the next step will be to equip artificial intelligence with predictive capabilities.
Based on previous vintages and the weather conditions of the year in question, the AI should end up being able to establish a prognosis on the qualities – and defects – of a product, even before it it could be tested by humans. These will lend themselves sooner or later to the game of tasting, which will then allow the algorithm to know if its predictions were correct – and to store experience, whatever the case.
Specialists warn against the perverse effects of such an invention, noting in particular that malicious people – yes, it exists – could manage to manipulate this kind of tool in order to “to rot an adversary and destroy his business financially”as Ben Zhao, an expert in machine learning and in cybersecurity at the University of Chicago.
The designers of the tool, on the other hand, want to be reassuring, highlighting the prodigious saving of time that can represent, for exhibitors and resellers, the effortless obtaining of original texts to use to promote their products. An unfailing positivity that stems from blindness or the most total ignorance of the workings of our internet.
We wish to thank the author of this post for this outstanding web content
This Robot Writes Perfect Wine Reviews Without Any Tasting
Visit our social media profiles along with other pages related to themhttps://www.ai-magazine.com/related-pages/