Writer's Guidelines Thank you for your interest in writing for AI EXPERT Magazine! AI EXPERT is the magazine of artificial intelligence in practice. Our features include case studies, how-to articles, technology tutorials, and product reviews. The focus: expert systems and neural networks, plus coverage of other AI topics like genetic algorithms, fuzzy logic, intelligent databases, and object-oriented programming. WHO ARE OUR READERS? AI EXPERT's readers are programmers and technical managers who have heard enough hype -- they're ready to put AI problem-solving techniques to work today. Our surveys show that about half our readers are developing expert-system or neural-network applications, although many more plan to begin such projects in the near future. They're looking for ideas and tools they can apply to their company's problems. They also want the details, technology tutorials, and product reviews which will help them implement those AI projects in the next year. AI EXPERT's readers fall into three main groups: those which program in traditional AI languages (like LISP or PROLOG), those using object-oriented languages (such as C++ or Smalltalk), and those using off-the-shelf expert-system or neural-network shells. Although we generally like to have our features be hardware-independent, you should note that three-quarters of our readers have chosen MS-DOS running on the Intel 80386 processor as their favored development platform. Even so, interest in other platforms including the Macintosh, UNIX (on several processors), IBM RS/6000, and VAX/VMS, runs high. Expert systems are the number-one topic, says our most recent reader survey, followed closely by development methodologies and case studies. Near the top of the list are neural networks, object-oriented programming, genetic algorithms, database management, fuzzy logic, and AI hardware. WHAT ARTICLES DO WE WANT? AI EXPERT runs four types of feature articles: case studies, industry overviews, how-to articles, and technology tutorials. Case studies examine a successfully deployed artificial intelligence project. We like to see plenty of detail, not only about what went right, but also what went wrong. Our readers take case histories and apply them to their own situations, so include specifics about which decisions were made, and why. What was the initial problem? What traditional solutions were tried, and why did they fail? How long was the development process, and what hardware and software tools were used? Who was involved, and for how long? What was learned during the pilot project? What was the final result of the project, and how much did it cost? What would be done differently next time? Those are only some of the questions you should answer in detail. Industry overviews are a special kind of case study, examining how AI technology is applied to various disciplines. How can neural networks benefit assembly lines? How are medical professionals using expert systems? Where are intelligent databases used in the financial community? Industry overviews require extensive contacts in the chosen field, so that a broad cross-section can be examined. How-to articles give plenty of hands-on examples of the nuts-and-bolts of AI development. Make your concepts independent of any particular hardware platform or development environment, except as part of your examples. How can you implement a neural network in C++? Explain the concepts, product annotated program code, and talk about why it works -- and what to do with it. Technology tutorials go into less depth, but over a broader area. What are genetic algorithms? Describe the latest developments and applications. Use plenty of examples. What's the best way to find an AI consultant? Talk about costs and benefits, the pros and cons. Get plenty of quotes, from vendors, consultants, and from satisfied and dissatisfied customers. HOW ABOUT REVIEWS? Writer's Guidelines, Page 2 Reviews examine a single new product, such as an expert- system shell, neural network development kit, or new AI- applicable language compiler. AI EXPERT also prints hardware reviews, such as of a neural network chip set or a high- powered numeric processor board. Reviewers must have no affiliation with the product's manufacturer or vendor and should have a genuine application for the product. Reviews must answer several questions, including: Does the product work as promoted and advertised? Is it free of substantive flaws? Where does it shine? What are its weaknesses? What should it be used for? What shouldn't it be used for? And, since no product is an island: Wow does it stack up against its direct competitors? Draw conclusions, and back them up. If the documentation is lousy, say so -- give examples. If the program is fastest in its class, say so -- give the test methodology and benchmarks. If a competing product does a better job -- say why. A review that merely describes the product isn't a review at all, but a waste of space and the reader's time. Product overviews look at more products (often between three and eight), but offer less detail about each product than would a stand-alone review. A good overview helps readers distinguish between competing products by highlighting key differences; a poor one runs page after page of boring feature-comparison charts, and makes the reader do all the analysis. (See the accompanying Reviewer's Guidelines for more detailed information.) WHAT'S THE WRITING STYLE? We have performed extensive research into subscriber demographics and reader/editorial relationships, which has led to the conclusion that the publication should adopt a stylistic methodology optimized for maximum readability coexisting with an emphasis on reducing obfuscation and excessive verbiage, towards the oft-desired but seldom- realized goal of reducing cerebral-tension syndrome. Or, to put it another way, our readers will have fewer headaches if you stick to an informal, easy-to-read style. AI EXPERT is a magazine for programmers and technical managers interested in applying AI technology today. It's not an academic journal citing the latest research which might find its way into applications in the year 2037. That's why we stress a conversational style. Pretend you're speaking with a group of friends or co- workers; that's the tone we want. The same is true of diagrams: Use all you want, as long as each helps explain the topic. Simple charts -- like those you'd draw on a blackboard or scribble on a napkin -- are the best. Writer's Guidelines, Page 3 Sidebars are an excellent way of handling information peripheral to your main topics, such as an historical perspective, a case study, or a related interview. AI EXPERT is a magazine, not a technical journal; it's not appropriate to cite every idea or concept formally with a footnote. That's why we encourage you to include a list of books or other printed works that you'd recommend to the interested readers. List the title and author, publisher, city, state, and date of publication, plus a couple of sentences describing the book's subject and intended audience. Regarding program listings: If you're teaching a concept, please use pseudocode so that you can explain the algorithm without confusing those readers not fluent in your chosen language. Presenting ready-to-run code? Please stick to standard language forms. When developing your article idea, make sure the main point is broad enough to be interesting to people working in a variety of hardware and software environments. Assume the reader has only limited exposure to the AI topic you're covering but is aware of relevant general principals. Finally, you're writing to your friends and colleagues. Don't write to impress but to educate. WHAT'S THE NEXT STEP? When your proposal is ready for submission, you can call or write. If you'd like to write to us, include a brief outline (and be specific!), along with a note explaining your credentials. If you have previously published writing samples, send along one or two; magazine samples are preferred over academic papers published in technical journals. Send your proposal to: AI EXPERT Miller Freeman Publications 500 Howard Street San Francisco, CA 94105 Be sure to include your daytime telephone number, fax number, and E-mail address (if you have one). Rather call and discuss your idea? No problem. Call Alan Zeichick, Editor, at (415) 267-7679, during regular west coast business hours. You can also reach Alan at CompuServe address 76703,756, or Internet address 76703.756@CompuServe.Com. Writer's Guidelines, Page 4 After we receive your outline, someone from the staff will call or write. We'll specify an article length and due dates. After the final article is received, we'll copy-edit the text according to our style guidelines. You'll receive author proofs for correction. Finally, when we receive the magazine back from the printer, we'll send you three complimentary copies, plus your author payment. Our general rates are $75 per full magazine page for reviews, $50 for articles. We look forward to working with you! Writer's Guidelines, Page 5 Reviewer's Guidelines So, you want to write a review for AI EXPERT? Great! Here's how you proceed: Talk to the magazine's editor, Alan Zeichick, before you begin working. You should ensure that the package has not already been reviewed or that a review isn't in progress. Be prepared to state your qualifications for the review. One qualification may be that you're an experienced user of the product, and you have already deployed at least one application using it. Another may be that you have a genuine use for the product. AI EXPERT will not use reviewers who have any ties to a product's vendor. We won't use beta testers, company employees, or anyone who might possibly benefit financially or personally from a positive review (or suffer from a negative one). We generally don't use reviewers recommended by the vendor for the same reason. When you've received your assignment, you may need to acquire the review product, if you don't have it already. Don't call the vendor yourself -- they're skeptical of such calls from freelancers and treat them with suspicion. Ask your editor to acquire the product for you. Often you will be permitted to keep the evaluation product. Note that sometimes the software you receive for review must be returned to AI EXPERT after the review is completed (generally because the vendor has requested its return). We will not issue author payments or subsequent article assignments, in those cases, unless all review software is returned. When writing the review, bear in mind that your job is to advise readers whether or not to purchase the product. The main criteria are: * Does the product perform according to the vendor's claims? * Is it a good value for the money? Your job is not to defend or attack the product or be concerned about the vendor's financial success or failure. Your job is to advise the potential buyer whether to buy the product, give it serious consideration, or walk away from it. A well-written review will briefly describe the product's intended use, with an example chosen from your own real-life test suite. However, a description alone does not a review make. Writer's Guidelines, Page 6 A comprehensive review will answer the following questions: * What are the product's hardware and software requirements? Are these requirements out of the ordinary, and if so, are they justified? (You should be harsh about products that require a full 640K of RAM and don't allow you to load your favorite TSR.) * What platform did you test the product on? How was performance from a speed and compatibility perspective? What would you recommend as the minimum hardware configuration for regular users of this product? (Truth in advertising: Very often the specification will read "Any IBM-compatible," but the snail-like speed mandates a 386.) Writer's Guidelines, Page 7 * Does the product conform to currently accepted hardware and software standards? Is it portable across platforms? Does it run on a network? Can its data be shared with other products? If not, how significant a weakness is this? (A broad-spectrum neural network should import data Lotus spreadsheets and dBase III files.) * How long did it take you to learn the product? What background must the average buyer have to learn the product? Is the learning curve especially long or short? What technical support does the vendor offer new users? Is there unlimited free 800-number service, or does the user pay per-call after 90 days? When you called the support line, what quality response did you get? (Bonus points for on-line tutorials, or optional two-day training seminars.) * Once learned, is the product easy to use? Does it grow with the user? Does performance bog down exponentially with the complexity of the program? What problems did you have? What bugs or "undocumented features" did you find? (If the maximum number of rules is 64, it's probably not for professionals.) * How good is the documentation? Is there plenty of it (maybe even too much)? Is it well indexed? Is it well written? Does it appear to be accurate? Are there sufficient illustrations? Is there a quick-reference card? Is the on-line help context-sensitive? (There's nothing worse than ordering an extra-cost tech reference manual when it contains essential information for using the product.) * How is the pricing schedule determined? Is there an evaluation policy or period? Is there a student discount? Are there site licenses available? Is there a demo version? How about a guaranteed maximum upgrade price? What's the warranty period? How does this compare to other major products in the marketplace? What's the company's track record here? (A money-back guarantee is a strong incentive to recommend that people try the product.) * What's unique about the product? What are its best and worst features? What obvious features is it lacking? How signficant are these pros and cons? (Be on the lookout for gimmicks that look good on the spec sheet but add little to the product's usefulness.) Writer's Guidelines, Page 8 * How does the product compare to others you've tried? Would you honestly recommend this product to your best friend? If you bought it yourself, are you glad? If you would recommend it to only some people, who are they? If some people should stay away from this, who are they? ("This program is good for students, but it is too slow and consumes too much memory to make it appropriate for serious expert-system developers.) BE SPECIFIC! Note that when you call the vendor for advice or to test the quality of their technical support, you should NOT identify yourself as a reviewer for AI EXPERT. Vendors lay down the red carpet for reviewers; by identifying yourself as one, you'll not be giving the reader an honest evaluation of the vendor's commitment to end-user support. If you follow these guidelines thoroughly, conscientiously, and honestly, you'll write an good, solid review. Thank you for your interest in AI EXPERT magazine! Revised 8/28/90 (a) Writer's Guidelines, Page 9