Write the reporter friends that AI stalks of grain, your understanding to artificial intelligence ma

Work of big data abstract, compile | Keiko, aileen

AI news is in rapid and occupational the head of each big door website and date of small letter public, large quantities of one AI run buccal reporter is corresponding also emerge in large numbers, the good and bad are intermingled.

NYU teachs Julian Togelius to be aimed at this brand-new news domain, an article was published on his rich guest, put forward to give reporters the proposal when writing article of concerned artificial intelligence.

After this text appears, the industry expert that includes Yann Lecun inside nods assist transmit in succession, be the one's innermost thoughts and feelings that spoke them really probably. And besides AI reporter, the fact that mentions in this article and proposal, the person that needs each care artificial intelligence probably is read carefully.

Author: Not, I am not the strange old man of a temperament what bad luck

Dear reporter friends,

The near future often can read many artificial intelligence relevant " sodden article " (news and public relations draft have) , reason wants to be aimed at it how be written is better to how be written realer artificial intelligence news raises a proposal a bit for everybody. Some articles basically are gibberish, some errors are full piece, some is like it seems that very rely on chart but also be full of misdirect. I won't designation path surname, any people that work in AI domain or often read AI news can have big case. Of course, also have a lot of very marvellous very good article, the scale of good nevertheless literary works still remains to rise.

Above all, I understand, what what you write is a domain that changes in quickness, was full of various professional term and the professional public figure that have great distant view among them. Besides these exciting things, this gets the content that still has a large number of is worth that write inside the country, but you do not understand too actually or say not to understand this domain completely.

Your understanding to artificial intelligence may is opposite with me " leather is made " understanding is euqally little.

But, the development process that leather makes is very slow, involve particular raw material and machine equipment. What artificial intelligence develops however is very rapid, and involve relevant content is you cannot lay a finger on perhaps sees. Your meeting generation must draw up one kind immediately before news outdated newest evolutional angst, but you do not know to begin proceed with from where however, decode the odd expression of one's political views of those artificial intelligence investigator. You want to draw up of course can be read by the masses the essay with self-identity and high hit, but you do not have too much time again, so this is not a relaxed work.

To be opposite so artificial intelligence domain has more the report of high quality, and be based on what I draw up to reporter of different technique actual strength cover the satisfactory rate of content, my general gives out to suggest reliably below.

Artificial intelligence is very big and the field with very wide branch

Fact: Artificial intelligence is a very big field, from relevant and topical will look with use method, be metabolic diversity.

The major conference of artificial intelligence domain (for instance IJCAI, AAAI, ICML and NIPS) have by tens of thousands the person that attend the meeting, among them great majority person understands the one fraction content in the conference only however, when I go attending these meetings, I can be understood only probably the content of 20% gets a few inspire. Although I myself am likely not quite clever, but can the branch domain with follow-up diversiform at the same time artificial intelligence, for instance the tie travels, the person of deepness study and the forward position knowledge that search randomly is very exiguous.

Proposal: Not acquiescent the student that you interview understands " artificial intelligence domain is foremost recently what is of the edge " .

More important is, what if someone says they know artificial intelligence domain is in now,producing, acquiesce please they understand the one fraction in this big field only. The investigator check of the other part in following AI domain please so one.

The system that does not have any artificial intelligence can make many tasks

Fact: Did not call somewhat at all " technology of a kind of artificial intelligence " exist.

Of the method that artificial intelligence is the software that development copies the mankind to use cerebrum to do a few businesses and concept close market. Investigator and developer development go new method (also use already some methods) go compose builds a few eye-catching software (it is hardware occasionally) , can play game to perhaps draw feline Mi for instance. Nevertheless, same the system cannot play game and picture cat Mi at the same time. In fact, I had not heard of the artificial intelligence system that can make a lot of tasks. The student with same even when is based on same idea, when developing a system to finish different job, they can choose to develop many software system to come true. Write when some reporters " the artificial intelligence technology of X company can have driven not only, still can write a poem now " when, these reporters blurred to actually another systems came true to write poetic fact, give a reader the illusion that a kind of machine that has special intelligence exists. The fact is not such.

Proposal: Do not say " technology of a kind of artificial intelligence " .

Go digging systematic restricted more. The nerve network model like going to those who see aerospace aggressor and unconscious Tesuma avenging these two game to apply (clew: Not be) .

What is artificial intelligence strange thing

Fact: Artificial intelligence is an old territory, and have a few ideas only among them be right new idea.

Nowadays very cruel dazzles but a little the nerve network of 889 time studies the century on the progress traceable that some are learned by the deepness of overmeasure, and nerve network model is to be based on more on the concept of 40 time mixes the century experiment. Below a lot of circumstances, forward position research is by right already methodical is tiny alter and improve composition, and when these already had a method to had not been born in investigator, had been come out by the design. " retrorse transmission " this algorithm is now of deepness study basically prop up, the history that has had several 10 years, and be discover independently by much person. The conquer of deep blue computer of IBM chess master Gairuikasipaluofu proved computer compares the mankind to be able to play chess more, among them core algorithm is Lan Tu having A clever the minimax that 1940 era puts forward is worth algorithm. Turing, artificial intelligence and one of author with broad sense scientific computer, the name was written to be 1950 " computational machine and intelligence " paper. The Turing that this paper weighs now because of us not only checks and famed, what also included the idea of a lot of core of artificial intelligence is budding.

Proposal: Read Turing the paper 1950. This is one reads the article that feels very marvellous, formulary without complex maths, all sorts of technology vocabularies are very common also blend in among them.

Your meeting exclaim has reached the designated position at that time at a few crucial thoughts of artificial intelligence, although the likelihood is budding phase only. In compose when the story about exciting technical new breakthrough, the proposal seeks advice to there already was a respectful form of address for an old person of some of age in this industry more, or the AI investigator that is middleaged person at least. These people become in AI very begin to study AI before cruel, had experienced the loop that an AI acclaims so. Likely this individual can tell you this new progress is old think of a way is improved a little from which a bit and come.

Do not believe hype, seek objective opinion more

Fact: Researcher always wants what to sell a dot.

Apparent, those people that work in poineering company are seeking the appraise cost that raises a company and the opportunity that invest or buy. The person that works in academia is searching a talk to invite, the article cites, promotion. Those people that work in big company hope people is certain to theirs concern as a result with real research (have nothing to do even) the product is interested.

Proposal: Do not believe hype.

Do not want what the relevant personnel of this article recommends you to look for compose interview a person, the attempt searchs other researcher and seek their opinion, whether does the inquiry believe the view of the description in this article.

Artificial intelligence comes from the wisdom at the mankind more, it is automation not completely

Fact: Major " artificial intelligence " the wisdom that is a person actually.

Researcher and development staff are dedicated at applying AI at specific domain (be like a robot, game or interpreter) the reason is: When compose builds a system to solve a problem, a large number of knowledge related to real problem (" domain knowledge " ) be included to be in the system. For simpler structural frame system, the likelihood needs to offer special input for the system, use the training data that prepares technically for its, artificial hand uses the code that has partial system, or need to be written afresh to solve a problem even.

Proposal: Understand " AI solution " what part is automatic, what is because the hand of domain knowledge moves the mankind to code and come,.

If be on a different question,a good method is query " , how will this system work " .

I am written first here, this slants the article may sound a little irratable. Actually, my disposition is not irratable, I also not old (I or a darling! ) . And I do not want to let a person feel AI does not have what progress to deserve a report.

In fact, already a lot of very fierce progress are worth follow-up, so we do not need to use all sorts of sale skill old hear regard as news will write. But at least should honest, take critically thinking and bona fide writes each article, be?